|back to main page|
Inazuma Mk.1 DRSSTC
Inazuma, from the Japanese for "lightning", is a mid-sized Dual Resonant Solid State Tesla Coil.
The initial run up provided 18"+ arcs to air with a 200V, 2A input from the variac with (IIRC) 100Hz and 150us on the interrupter. This was "interrupted" by all four IGBTs going *pop*. A short video will follow.
I'm not sure what caused it, but I've replaced the bridge connections with a lower inductance layout, thicker copper connections and double the HF decoupling. Farnell also ran out of the ST IGBTs I was using so I had to get some HGT40N60 replacements instead.
I've wanted to build a DRSSTC for a while now and seeing several on t'internet didn't help. However, lack of easy + cheap access to IGBT bricks and even mini-bricks led me down the route of choosing to use TO-247 IGBTs instead.
These are miles cheaper, just harder to mount properly as the drains are connected to the tab which causes the heatsink to operate at the drain potential. This means that insulation has to be placed between the device and heatsink or separate heatsinks have to be used for each device (this is the route I've gone down).
Partly because I've gone for smaller devices that can support less power dissipation and partly because I've had good times with his other circuts, I've copied Steve Conner's PLL driver for DRSSTCs for this coil. This offers control over the phase of the switching waveform as well as providing deadtime control.
This means that I can set up the coil with the minimum of stress on the swiching devices. The tradeoff is a much more complex control circuit than more common DRSSTC circuits but building the electronics is the bit that I like the most!
A piece of gas supply yellow (HDPE?) pipe of 125mm diameter, wound with 0.25 mm Grade 2 Enamelled Copper Wire (ECW). Grey plastic end caps were fitted to the top and bottom and attached with nylon countersunk machine screws fixed into holes tapped into the pipe. Several coats of Ronseal Hard Glaze polyurethane varnish were applied for mechanical protection and added electrical insulation.
I made an attempt at making a motorised jig to aid winding. It used a very old Black and Decker cordless drill run from a variable supply with a model gearwheel attached to a shaft in the chuck. This engaged with a model gearwheel glued to one of the secondary endcaps. The secondary rotated around a piece of M4 threaded rod.
This worked fine for the first 2cm of turns, the power supply was not quite man enough to turn the former with me resting my finger on it but was enough to keep the wire under tension - power assisted winding. However the glue I'd bodged the gear onto the shaft with kept coming unstuck so I gave up trying to fix the jig and wound the coil by hand. Suprisingly, it only took about 2 1/2 hours to wind by hand. The motor came in handy for keeping the coil moving when applying the varnish.
I used 100mm diameter aluminium ducting from B+Q wrapped around a 150mm perspex former and hot melt glued in place. The join and the central former were covered in aluminium tape.
Primary, Geometry + Resonant Capacitor
Primary / secondary geometry was calculated using mandk to produce a coupling of around 0.2. The aim was to increase space between primary and secondary to reduce the risk of significant flashovers. Mandk is a bit awkward to use but it confirms the results in my spreadsheet and is much better than guessing!
The primary supports were made from a polypropylene chopping board (Sainsbury's
I made some longer thinner rectangles to hold the primary strike rail, driled through and countersunk them. I then tapped the primary supports and fixed them together with a little bit of hot melt and nylon bolts. It looks a bit scruffy but it more than does the job. I then used plastic right angle brackets to secure the primary mounts to the baseboard.
I bought a load of yellow 470nF, 1000V, high pulse current rated, plastic film, axial capacitors for the Multiple Miniature Capacitor (MMC) bank. I had orginally designed a 300nF, 4kV capacitor bank, but on the advice of the good folks at 4hv, I decided to use a lower value. I settled on 2 parallel strings of 9 capacitors to give me a 104nF, 9kV capacitor. I've built the bank with the ability to tap the string at 9, 7 and 5 caps so I can up the power if necessary.
The base unit was made from a length of 300mm deep pine shelf cut into two squares. Four lengths of M10 threaded rod and bolts were used to support the upper part of the base upon which the primary and secondary are mounted.
Full bridge made up of four ST STGW30NC60WD TO-247 packaged IGBTs (600V, current 60A, pulsed 200A, Vce-sat = 2.2V). These are packaged with an anti-parallel ultra-fast diode (Vf = 1.5V, trr = 44ns). These were selected because at the time of building, they were a mere £2.79 from Farnell (part # 129-3659). Compare this with the price of a 40N60 mini brick @ £26.33 - almost 10 times more!
I designed a PCB to hold the four TO-247 devices in the full bridge configuration with a laminar bus structure to reduce the supply inductance. The input decoupling was provided by a low ESR / ESL 3300uf, 420V electrolytic made by Rifa with the connections to the PCB made with copper foil. There was also a 470nF, 1kV film capacitor to provide local HF decoupling. Because the PCB was only made with 1oz copper, I reinforced the primary current paths with copper foil and copper de-solder braid to reduce the resistance.
Each IGBT was clamped to a multi fin CPU heatsink using a short length of aluminium U-channel screwed into M4 tapped holes in the heatsink. The M4 machine screws were fitted with spring washers. These measures provide a greater degree of positive pressure between transistor and heatsink than just using the mounting holes, which reduces the thermal impedance between device and heatsink.
A single heatink per IGBT was used to keep the drains isolated from each other - this saves on having to use electrically insulating materials which increase thermal impedance between heatsink and device junction.
Four Adda, 60mm, 12V fans were used, two blowing and two sucking to move air over the heatsinks and bridge assembly. Fan speed control was provided by a PWM circuit based on this one, and the fan voltage was adjustable over the rage 5V to 12V using a panel mount pot on the controller. Switching of the fans was achieved by using a IRF1010 N-channel MOSFET. Some "growling" was observed at low duty cycles, but that will be quickly masked by the noise from the streamer.
Mains rectification was proided by a single Fairchild GBPC3508 800V bridge rectifier.
I've chosen Steve Conner's PLL controller for reasons outlined above and I'm leaving it essentially unchanged.
I've left off Steve's metering circuit in favour of doing a bargraph type meter: however I'm too tight to buy an LM3914 meter IC so will probably do something with a few LM3339 comparators and a load of resistors.
The current transformer uses and identical core and wire as my GDT (see below) with 47 turns on the secondary and a one turn primary. The burden was formed by 10 * 4R7, 1/4W metal film resistors in parallel giving 0.47 ohms total resistance and a voltage output of 10mV/A.
As in Steve's controller, this single transformer is used for both feedback and over current protection, with the feedback taken across the burden in series with a couple of anti-parallel conencted EGP20D diodes. This increases the lock sensitivity of the PLL circuit at low primary currents.
Back of the envelope calculations show that the resistor will dissipate about 2W worst case (max primary current and duty cycle) so 2.5W of resistor should be more than adequate.
All of the components were shoehorned into an aluminium diecast metal case to provide some shielding from the effects of the coil.
There are holes in both lid and sides of case to allow the adjustment of gain, phase, frequency and deadtime controls wthout the need for dissasembly. This required the drilling of holes in the PLL circuit board to allow access to the screw slot on the underside of the pot.
The current limt pot was a panel mounted type with a knob to allow easy adjustment of the limit set point.
Gate drive used two UCC37322s with outputs clamped with 1N5819 diodes to the rails, 2000uF worth of low impedance electrolytic and 400nF of ceramic capacitor decoupling on the 15V supply rails. Drive coupled to the primary through 4uF of film capacitor.
A snapshot from the spreadsheet I use to design my GDTs with is shown below. I ended up using 0.5mm diameter triple-insulated Furakawa wire to get good isolation between primary and secondaries. All of the secondaries are wound on the same core. I used an identical core for my current transformer.
Duty cycle can be adjusted by using the deadtime control so that the pulse width ranges from 50% to about 30%. This deadtime means that one IGBT in a leg of the full bridge will not turn on when the other IGBTs "tail current" is still present. This will prevent too much heating of the device and hopefully prevent damage.
The interrupter is based on two 555 circuits: an astable for the pulse frequency which triggers a monostable for the pulse width.
Frequency range = 0.2Hz to 225Hz
The SPST momentary "fire" button connects the 9V PP3 battery to the circuit to prevent any false triggering. An LED is lit when the button is pressed to cleary indicate activity.
To provide electrical isolation from the coil, a SPDIF / TOSLINK fibre optic emitter was originally used as the output. This was connected via a 3m cable to the main controller. However some problems were encountered with the SPDIF emitter.
Firstly, the SPDIF receiver I bought from Farnell appeared to have an AC coupled output in that it gave a short low > high > low pulse on the rising edge of the input pulse and an identical pulse on the falling edge. I could have fed this into a toggle latch to provide the pulse on the output but this presented two problems
In the end, I went to a photodiode receiver and an LED emitter. The LED is a super high bright AlGaAs 5mm packaged diode with a 20 degree focused beam. This was mounted in a modified SPDIF emitter package with the back drilled out and the LED held in place with hot melt and driven at 25mA.
I initially tried a phototransistor as the receiver but it's response to the pulses was far too slow. I ended up using an SFH213 photodiode mounted in a SPDIF receiver in a similar fashion to the LED emitter. I tried to buffer this using a TL082 in a zero bias circuit but was getting odd results.
A work colleague pointed out that it was supposed to be a dual rail op-amp and Steve Conner pointed out that the input range went as far as the input rail but not down to the negative rail as I wanted. I swapped to a good old LM358 and my problems were solved. I ended up using the other half of the op-amp as a comparator to get the pulse width output as required. The slew rate is a bit slow but it's fine after being buffered by a transistor to get it up to 15V logic levels to be input into the controller.
The entire assembly was shoe-horned into a small plastic case with a front panel designed using Visio.
Old stuff from work
DRSSTC Information & Further reading
These are links to the information I found most useful when designing and building this coil.
Community Wiki - DRSSTC