Self-oscillating teslacoil

Abstract

The Idea is quite simple: Every cheap energy-saving lamp has a self-resonant voltage inverter inside. They are designed for low-power operation up to a few watts. Why don't scale up the whole thing and replace the resonance circuit for generating the needed lamp voltage with the tuned primary of a double-resonant solid state tesla coil (DRSSTC)?
The guys from teslacoil.net did this already and I think that the SSTC3, SSTC3.7, SSTC3.8 and SSTC3.9 as well as some of the other coils are based on such a circuit.
My experiments showed that the idea is good and you can create amazing lightning with a very simple and reliable circuit. The only thing you need - apart from the simple self-resonant power stage - is a suitable power supply with a current-limiting feature.
But why stop thinking here? We could even try to avoid other major disadvantages that a DRSSTC has: Sound intensity and relatively thin discharges, especially at low-power. The discharges from a interrupted solid-state tesla coil (ISSTC) are more pleasant to the ear than the screaming noise of a DRSSTC at high BPS and somewhat brighter. This is because of the long pulse time with a sinusoidal waveform (that is derived from the mains) used in a ISSTC.
Regarding DRSSTCs with long pulse times, there was some discussion on the web a few years ago, but I could not figure out where this was. However, as I think that there are currently no circuits like the sketched one in the public domain, I decided to spend some time on developing such a circuit: A simple, self-resonant DRSSTC with long on time and sinusoidal pulse envelope.

The circuit

Following the often used name convention, this coil may be called a self-oscillating double resonant solid state tesla coil (SDRSSTC).
It consists of a power supply part with the pulse generator and a self-oscillating power inverter.

Power supply

As the waveform of a self-oscillating tesla coil can't be controlled in the power stage, this control has to be done elsewhere. The power supply was chosen to be the right place for this (see Figure 1.1). TR2, D10 - D13 and the circuit around IC7 and IC8 are the low voltage power supply for the control logic. The five volt regulator (IC8) is a spare part for future use, where the voltage may be needed.
The toroidal inductor L1 has a half-turn primary that carries the primary resonance current. The output current from L1 leads to a voltage drop at R26, which is rectified by D15 - D18. C9 smoothes the signal. The following comparator (IC3) generate a logic high signal on it's output if the current threshold is exceeded. The current threshold can be set using R23.
For the zero-cross detection, the secondary voltage from the mains transformer is fed via D9 or D21 and R10 into the base of T3. The placement of D9 or D21 can be used to select the right phase, only place one of the diodes at a time. The delay time between the pulses is generated by the circuit around IC1B and synchronized with IC2A and IC1A. The over-current shutdown release is synchronized too, using IC2B. The over-current shutdown can be triggered at any time and force the output low using T7. This transistor is connected in a wired-or configuration with T6 to generate the power switch signal. This signal is high for the duration of one halfwave period of the mains voltage and switches a MOSFET in the power stage supply line on, until an over-current condition occurs.
The rest of the circuit, consisting of T4, T5, IC1C and IC1D, generates a pulse burst for triggering a SCR that switches the power to the power stage. This option has been used when running the circuit without over-current protection, as SCRs have been found to be more robust than the MOSFET switch.

Power stage

The schematic of the power stage is shown in figure 1.2. The IGBTs T1 and T2 form a half-bridge circuit. They are protected by D1 - D6 against voltage transients and reverse current flow. R1 and R2 damp the oscillations on the gates of T1 and T2. L2 and L3 are toroidal inductors and are used for the feedback. A small amount of the output power is coupled back by the two turns through both feedback coils (and the current measure coil, see figure 1.3). D1 and D2 are suppressor diodes that limit the voltage on each gate to +-15 volts.

Figure 1.2: Power stage

Figure 1.3: Feedback scheme

Starting circuit

The starting circuit (left side of figure 1.2) consists of C1, D7, D8, R3 - R5 and SCR1. C1 is charged through R3 and R4 when power is applied to the power stage. When the voltage level of C1 reaches approximately 30 volts, D8 triggers SCR1, which discharges C1 into the gate circuit of the lower IGBT (T1). The following current pulse in the primary resonator starts the oscillation. R6 and R7 ensure that the resonance capacitor is pre-charged before the starting pulse is triggered. In normal operation, C1 is discharged through D7 every time T1 turns on, disabling the starting circuit. If the starting process was not successful, C1 is charged again and the process repeats.

Assembly

Primary capacitor

The primary capacitor is made of four WIMA FKP-1 capacitors in parallel, 10 nF / 1600 V DC each. This gives a total capacitance of 40 nF at 1600 V DC.

Primary coil

The primary coil is made of 5 turns PVC isolated stranded wire, 4 mm (0.157") in diameter. It is wound on a PVC former that is 16 cm (6.3") in diameter. The primary coil is elevated 6.5 cm (2.56") above the secondary base.

Secondary coil

The secondary is wound on a PVC pipe with a diameter of 10 cm (4") and a length of 45 cm (18"). There are 1140 turns of 0.3 mm (awg 28) enameled copper wire. The calculated resonance frequency is approx. 350 KHz, but the operation frequency is set to around 300 KHz which seems to match the secondary resonance frequency pretty good.
A breakout point is added on the top of the coil to guide the discharges upwards.

Development results

Impact of switching delays

In an ideal dual-resonant tesla coil circuit, the switching of the output voltage of the bridge occurs exactly when the current in the primary circuit crosses zero. This will avoid switching losses, as there is no power flowing when the IGBTs switch. In a real circuit, there is always a small delay or phase shift between the zero crossing of the current and the switching action. This phase shift increases the losses and the stress to the IGBTs, so it is important to minimize it.
A way to deal with this would be to employ a PLL circuit that gets the bridge output voltage and the primary current on its inputs and tries to eliminates any phase shift between the two signals by changing it's output frequency. But for this self-resonant approach, there is no place for a PLL and therefore the only thing that can be done is the reduction of the phase shift.

Tips for the reproduction

Keep the feedback inductance low!

Measurements showed a considerable delay between the primary current and the gate voltage of the IGBTs, if the inductance of the feedback coils is too large. This behaviour was also found in simulations with pSpice. Minimizing the inductance helps keeping the switching delay and thus the switching losses in the IGBTs low. This becomes especially important if the power of the coil is increased. Of course, the transformation ratio and feedback power must be made high enough that the coil can start to oscillate.
A word about the feedback in general: One could think of a big power loss in the suppressor diodes, because there is no current limiting resistor between the feedback coil and the diodes. As we need a square waveform with fast slopes, there is a lot of power supplied to drive the gate voltage from -15 to +15 volts in a few hundred nanoseconds. As the voltage is going to rise further, the suppressor diodes start conducting and limit the voltage to +-15 volts. Fortunately, the small cores of the feedback inductors saturate quite fast then, dropping the inductance to almost zero. This means that there is no more energy transfer between the primary resonance circuit and the feedback circuits. It is important that the cores have low losses when going into saturation, otherwise they will become very hot. Good results were achieved using energy storage inductors for boost converters.
Do not try to add a resistor between the feedback coils and the suppressor diodes! This will prevent the core saturation, waste serious amounts of power and heat up the resistor and the inductors quite fast while limiting the primary current. One of the secrets is not to use any resistors there.

Keep the resonance frequency low!

The tesla coil from teslacoil.net has a similar height of 45 cm (18"), but uses 0.2 mm (awg 32) wire. Together with a bigger top capacitance, the calculated resonance frequency is about 228 KHz and therefore 35 percent lower than in the discussed coil. This helps keeping the power losses in the IGBTs low. The reasons are that there are less switching actions per second and that the delay between zero-crossing of the primary current and the actual switching action is a smaller fraction of the switching period if the operating frequency is lower.

Use fast IGBTs!

The used IRG4PC50W from International Rectifier are rated for "warp switching speed". The datasheet says that more than 300 KHz in soft switching (resonant) mode are possible. However, low switching times and delays are important for efficient operation. The switching times (rise and fall times) determines the amount of switching losses during each cycle. The switching delay will be added to the total phase delay between primary current and output voltage of the bridge and therefore increase the losses, too.

Use a current limiter!

The operating current can be set to a safe level for operation at ideal conditions. But every disorder of this operation (for example ground strikes) can lead to unpredictable current spikes and blow the output stage. And of course, the spark length should not be limited by high safety margins. So it is advisable to use a current limiter that switches off the coil for the current cycle if an over-current in the primary circuit is detected. The adding of an over-current protection reduced the quantity of dead silicon during development and operation dramatically.

Operation

Finally, there are some pictures of the coil operating.

Discharge pictures

The pictures are taken in front of the aluminium shielding that protects the surrounding equipment.

Figure 1.4: Discharges

Figure 1.5: Discharges

Further development

The next steps will be:

• reducing the phase shift between the output current and -voltage of the power stage as far as possible


• winding a new secondary with 0.2 mm (awg 32) wire and adding a bigger topload to reduce the operating frequency


• expanding the power stage to a full-bridge

Some of the changes are already made:
Self-oscillating teslacoil (extended to full-bridge operation)

Further readings

There is a lot of information already on the web:

easternvoltageresearch.com

Steve Conner's amazing tesla coils

Homepage of Steve Ward who built very powerful tesla coils

Derek Woodroffe's tesla coils

Home of the JavaScript Tesla Coil Designer

teslacoil.net   (Link is actually dead, hope it'll be back soon)