Pyramider: The evolution from Spark Gap to IGBT

A number of years ago I was approached about a Tesla coil in a deceased estate of a friend of a friend you might say. Naturally this was an opportunity too good to pass up.

Pyramider circa 2006.

Upon first inspection I found out the coil had three 12kV neon sign transformers and a rotary spark gap.


I duly agreed to take responsibility for the creation and it then sat at work for a few years occasionally being tinkered with.

Around 2011 He who must not be named on the internet moved back to little ole NZ bringing back with him enthusiasm for building tesla coils and a REALLY big piece of power silicon: Thus began my renewed interest in power electronics...

2.5kv at 1800A

The idea was to build what's called an Off-Line Tesla Coil (OLTC), in that there is no supply transformer: The incoming mains supply is rectified and fed directly into the tank capacitor.

OLTCs have some interesting design criteria: With the primary voltage being very low (compared to spark gap coil) The tank capacitor has to be huge in order to supply sufficient energy to the system.
The huge tank capacitor in turn means that the primary inductance has to be tiny for the system to have a sensible resonant frequency.

The primary currents involved are pretty massive, as a result the "wiring" has to be pretty heavy duty: In this case the conductors are made from 1.6mm copper sheet, 190mm wide. That is 300 square millimetres of copper!

Initial back of a envelope calculations were based on an input power of around 5 to 10kw to make things reasonably interesting. With a reasonably safe maximum switching frequency of around 7-800Hz this meant the tank capacitor needed to store around 10 to 12 Joules. Charging the capacitor to around 1KV, (more on this later) around 20µF was necessary.

Primary Construction:
As mentioned earlier the primary conductors are made from copper strip 190mm wide, there are a couple of reasons for this besides the ludicrous discharge currents...
The tiny primary inductance: Total inductance, which includes the parasitic inductances of the conductors, transistor, ESL of the capacitors etc had to be about 0.1µH or less. This is so that the resonant frequency of the primary matched the secondary.
The wider a conductor is the lower its inductance, therefore transmission lines made of nice wide strips of copper have lower parasitic inductances.
Flat conductors are also a lot easier to mount closer together: This reduces the parasitic inductance of the system further still. (more on this later)

The "Snorkel"Primary:
Owing to the low inductance needed, the primary had to be very small in diameter (the diameter actually ended up less than the height) this resulted in much headscratching... The required diameter was less than the diameter of the secondary... Eventually the solution of internally mounting the primary was devised by use of a little trickery:

In order to minimise parasitic inductance, I had to devise a suitable feed line to connect the primary turn to the capacitor and transistor assembly. A bit of a hunch and a few hours of sketching later a plan was devised to fabricate a shape from copper strip and then roll it into a combined primary turn and feed line... Below is a sketch of the result:

Sketch of the fabricated primary.

The above shape was made by cutting strips of copper sheet that were then TIG welded together, this approach was taken as there was a shortage of material (it was rather expensive)

The welds were linished flat and all the sharp edges were polished off. I then put the whole part through a sheet metal roller to give the final shape.

The many, many bolts used to connect
the snorkel to the capacitor assembly.

Snorkel featuring polyethylene
sheet to keep the feed line separated. 

Tank Capacitor:

Capacitors in Tesla coils have a pretty hard life, they're charged to high voltages and discharged very quickly... Only to be told to do it many times per second. So naturally they have to be very low loss and rated to huge pulse currents.

Capacitor plate to the left.

Currently about the best off the shelf solution is to connect multiple smaller capacitors in series-parallel to give the appropriate capacitance, voltage and current rating. In this case depending on tuning there are about 40 capacitors in parallel. The capacitors are mounted to a parallel plate feed line (think double sided PCB) made from two sheets of 1.6mm copper spaced 5mm apart. This is an idea I borrowed from Steve Connor and Greg Leyh. The capacitors connect to the plates using threaded brass studs and copper tubing spacers.

Brass studs and copper spacers fitted.

One day I would like to build a coil using water cooled capacitors designed for induction heating from a supplier such as Celem: Megawatts anyone?

Capacitors fitted to the assembly.

Top and bottom mounting bushes.

Mechanical Assembly:
Making the components necessary to hold everything together is a little interesting:
For one tesla coils generate high voltages - Many of the components have to be very good insulators.
Ferrous components can get hot due to induction heating (Tesla coils tend to generate rather large EM fields around the primary)
The result of all this is that I tried to use plastic components as much as possible...

Bottom mounting bush and threaded
nylon rod.

The four mounting rods.

The secondary former is supported by a combination of 10mm threaded nylon rod, 3d printed bushings and PVC pipe columns.

Columns and top bushing fitted.

Secondary former sitting on the bushings.

Water Cooling System:
For a lot of the initial running, the power consumption was limited to around 5kw. This changed quite dramatically when I finished building a three phase input supply. A brainwave to do with the implementing a pulse forming in the charging circuit made things even more dramatic.

The side effect of much more power is well, much more heat... Until this point we had been using a fan cooled aluminium heatsink, this was ok but things were getting a little hot.

The shiny new waterblock.

The transistor with heat sink.

Enter water cooling: I have a friend who's a rather brilliant machinist. We spent an evening drawing up a water cooling block to suit the transistor. An hour or so on the CNC machine later and presto.

Bolted together.

Getting ready to bolt the transistor to the waterblock: Toons,
safety glasses, and two finished cups of coffee... Serious work.

Nobody breathe.

For the first test I lashed the system together using a plastic bucket and a spare PC water cooling radiator I had lying around.

Hose restraint.

The first results were more than a little exciting.... The cooling water stayed below 27 degrees celsius during many power runs like the video below... Making Toccata and Fugue even more epic.

Zappy.

So What's Next??
I've made a better water reservoir than a bucket, Hansen plumbing fittings are great for building crazy stuff like this.

10 Litre reservoir.

Still to be fitted is a new radiator from Koolance: As a final teaser here's the CNC machined mounting brackets for it... Next post will detail the power system.

Problem: "Radiator bracket for water cooled Tesla coil"
is not a part that exists...
Solution: CNC machine.

Project Sigma Mesa: Achieving high speed imagery of exploding wires.

Sometime last year during a Friday afternoon session at Fidels Cafe. Jez, He who must not be named on the internet and myself were discussing the idea of doing some kind of project to present at the Wellington Lux symposium.

Due to quite an interest in things that go bang in an entertaining manner we came up with the idea of doing something to do with imaging exploding wires at the very very early stages of a high current pulse flowing through a fine wire. That is in a nut shell taking a photo of a wire just as it's being turned into plasma.

Most exploding wire setups involve the following:

Some kind of energy storage device, often a pulse rated capacitor.

A fast energy switching device such as a triggered spark gap or an SCR.

Some kind of power supply for charging the energy storage capacitor.

Below is one of our first images using this method of imaging an exploding wire, the total exposure time is 3.2 seconds, the wire was exploded using an Information Unlimited High Energy Pulser

The justification for this project is that a lot of existing exploding wire imagery has been done which reveals late stage explosion information, that is most images show spectacular showers of sparks. This is a result of using the method of opening the camera shutter, exploding the wire or other device under test, and then closing the camera shutter. Peter Terren, Mike Harrison and The University of Canterbury  have all done some pretty cool experiments in this area.

Now from our discussions we set out to use shutter speeds in the order of 1/8000th of a second as this is the fastest shutter speed my DSLR is capable of. The reasoning behind the use of super fast shutter speeds is to minimise motion blurring of the rapidly expanding plasma.

Due to the shutter speed being ludicrously fast, the timing of camera shutter and electrical impulse have to be very precisely timed in order to capture an image.

Spark Timing:

After the initial test using the machine with no pity, I set about doing much reading on very precise and repeatable pulsed energy systems. This paper proved to be very enlightening on the subject of low jitter spark gaps. The book: Electronic Flash, Strobe by Doc Edgerton also proved to be a gold mine of information on rapid photography.

I decided to build a field distortion gap from components salvaged out of the machine with no pity. Rebuilding the gap was necessary on as the existing one was mechanically quite flexible and thus prone to the electrode spacing changing.

The remains of the original sparkgap assembly

Spark Gap Construction:

The gap electrodes are 10mm solid tungsten bar which is clamped into aluminium cable lugs.

Two cable lugs are securely bolted together at 90 degrees: The larger lug clamps the tungsten electrode, while the smaller one is there to allow the conductor to be easily connected.

The spark gap baseplate is made from HDPE chopping board as it's a cheap HV tolerant construction material. Two small rectangular recesses were milled into the base plate to accommodate the smaller of the two cable lugs. 

This recesses ensure that the electrodes are kept aligned to each other and mounted rigidly to the base plate.

The trigger electrode is mounted using a single bolt as it is subjected to far less force in the course of setup and use of the system.

10mm tungsten is a fraction excessive for the trigger electrode however it is what I had to hand.

The back of the spark gap baseplate

A key feature of a field distortion spark gap is that the trigger electrode is biassed to half the voltage across the main electrodes. This was achieved by fitting 4x 10MΩ resistors as a voltage divider between the three electrodes.

The arrangement of the resistors also resulted in the sparkgap being nicknamed the flux capacitor of the system.

Spark Gap Trigger Generator:

In order to trigger the sparkgap an impulse of around 10 to 20kV is needed. The trigger pulse generator needs to have as short a propagation delay as possible with minimal jitter. I settled on a design using a salvaged pulse transformer triggered by a reasonably beefy SCR.

The 9v battery provides a supply to illuminate the LED of the optocoupler

The logic level of the trigger system is isolated from the high voltage side by a triac optocoupler.

After putting it together a quick test with the oscilloscope revealed a logic to spark delay of about 3.4µs and very little jitter.

Trigger generator on the test bench

Top trace is the signal
Lower trace is the spark current

The trigger impulse is capacitively coupled to the trigger electrode by two 1nF ceramic capacitors in series.

1nF capacitors connected to the trigger electrode

The capacitors are necessary to provide DC blocking between the trigger system and the energy storage capacitor. This prevents weird leakage currents showing up in places where they're not wanted.

Capacitor Charging Supply:
The power supply for charging the capacitor is a pretty simple affair, consisting of two 110-6500V transformers with their primaries in series and secondaries in parallel. (NZ has a 230v mains supply) The output of the transformers is rectified by four HV diodes in a bridge configuration. A 16kΩ wire wound resistor serves to limit the charging current and ensure there is no explosion inside the case should the diodes go short circuit...

The charge voltage is controlled by  feeding the supply from a variac and watching the voltmeter.

KillerVolts

Victim connection area

Overall Assembly of the System:

The sparkgap, capacitor and piece of fine wire to be exploded are all connected together in series with 6 square millimetre speaker cable.

No one touch anything.

Timing:
The propagation delay of the DSLR is such, that it must be triggered around 63 milliseconds before the wire is exploded. This timing delay is handled with the use of a Camera Axe running a custom firmware. Final calibration of the system was achieved by taking many photographs of the trigger spark at faster and faster shutter speeds.

Camera Axe lower left.

Testing revealed that the propagation delay in the camera varies by as much as 120µs depending what lens is fitted to the camera. Changing from landscape to portrait orientation increases the delay by 150µs... Enough to cause much hair loss over time...

Boom!

The pictures mostly speak for themselves... Here are some of the first shots that were successfully taken, All shots other than the light bulb are 0.15mm diameter copper wire being exploded at approximately 4000v:

The blue band in the image to the right is the rather bright flash caused by the electrical discharge: Things happen so quickly that in the time taken for the shutter to scan the sensor, quite a lot happens.

Rough calculations revealed that the plasma is moving at an average speed of around mach 2 during the duration of the exposure. The rapid expansion of the plasma is evidenced in the photograph to the left of a light bulb having a very bad day.

Changing the camera to portrait orientation so that the shutter scanned along the length of the wire revealed quite a lot more information about what was going on. Along with the fun of discovering the change in propagation delay of the camera.

At this point I turned my attention back to one of the original goals of the project: This was to create lettering with exploding wire plasma in camera.

Attempts at shaping the wire yielded no real results due to the plasma expanding far too rapidly. One interesting side effect was the result of what appears to be the shock waves converging in the corner of the wire: The corner was very neatly "cut" into the black foam core used as the background.

An attempt at creating the letter "L"

The "L" cut into the background.

The next idea was to constrain the plasma in some for of channel, long enough to take the exposure.
An hour or so in front of Sketchup later I had something on the go in my Ultimaker printer:

"L" just before being exploded.

The first revision of the lettering turned out to be too small. The "L" exploded violently whilst spraying PLA shrapnel all over the workshop.

The "L" exploding.

The "L" also failed to constrain the plasma for long enough.

Remains of the "L" mould.

After redesigning the moulds to be bigger and stronger, we started getting much better results:

One can see the wire fragmenting in the flash bar.