High Frequency Solid State Tesla Coil

HF-SSTC generating plasma flame

Introduction

HF-SSTC stands for High Frequency Solid State Tesla Coil. It is a small CW Tesla coil that operates at the unusually high frequency of 4MHz. This page documents the design, build and testing of a 500W demonstration unit. This page shows that there are significant challenges associated with this high operating frequency but it yields some interesting results and a relatively simple and elegant design. The HF-SSTC makes use of a standard power MOSFET in a modern switching amplifier design to produce the required RF power and drive the resonator to produce sparks. The information that follows assumes a basic understanding of how a CW tesla coil operates and some knowledge of power electronics and RF engineering.

 

Why build a Solid State Tesla Coil to operate at such a high frequency ???

Well there were several things that originally motivated me to do this:

1. To prove that it could be done, despite many technical hurdles to be overcome.
2. To see how the spark appearance and behaviour differ at MHz frequencies.
3. To make the system compact and portable. This implies a high resonant frequency.

I was initially inspired by a commercial project I was working on involving Class E RF power amplifiers. I thought that I could borrow some of the clever and relatively new techniques used by RF engineers and use them to make some sparks. I must also say that the success of Derek Woodroffe's MillieTess project also motivated me to do this project. Technical discussions with Derek helped me immensely.

I chose the operating frequency of 4 MHz for several reasons. Firstly it is close to the 80M amateur radio band. There are several designs presented for RF power amplifiers operating in this part of the frequency spectrum. I thought that these would provide a good starting point, and a resource to fall back on if I got stuck. Secondly it is close to the resonant frequency of Derek's MillieTess coil and his coil seemed to work fine.

Finally I thought that 4MHz was sufficiently far into the HF spectrum that I would not be able to use conventional power electronics techniques to develop the power required for impressive sparks. Therefore I would have to tackle this project from an RF Engineering viewpoint and this was a new and interesting challenge for me.

At Teslathons over the years several people had asked me whether an HF Radio Transmitter could be used to drive a Tesla Coil and produce sparks. I was almost sure that you could produce sparks, but with concerns about matching and VSWR, I thought it was time to try it and find out exactly where those sparks would come from!

 

The Tesla Resonator

I started this project by winding the Tesla resonator first. One of the original aims was to make this SSTC compact and portable for demonstrations. It was intended that it would be fairly tame and well behaved, but at the same time powerful enough to make an impressive display of corona and light fluorescent lamps etc. For this reason I chose the cardboard tube from the inside of a toilet roll as the former for the TC resonator. This has a diameter of 1.8" and a height of 4.3". I initially took a guess at the number of turns and wound it with 105 turns of 0.8mm enameled copper wire that I had handy. Sweeping this coil with a function generator revealed that it resonated at almost 8MHz.

4MHz Tesla Resonator

I thought that this frequency was a bit too high and that it might present problems for the electronics producing sufficient power at 8MHz. Therefore I doubled the number of turns and rewound the coil with 210 turns of Green "wire-wrapping" wire. This has a single 0.25mm diameter solid copper conductor that is silver plated and covered with Kynar insulation. The insulation is incredibly tough and resistant to scratching as this wire is intended for use in prototyping applications where it may be stretched over sharp metal objects, etc to make connections.

At 4MHz the skin depth in copper is only 0.033mm. The 0.125mm radius of the chosen wire represents 3.8 skin depths. Therefore it is my opinion that there would be little benefit in going to a thicker guage of wire, as the largest portion of the current is already flowing in the outermost skin of the wire.

I also believe that the silver plating of the copper conductor may provide a lower resistance path to the majority of the current that flows on the surface of the wire at these frequencies. This may reduce the AC resistance slightly and therefore reduce the amount of heating due to power lost in the winding resistance of the coil. This may become an issue if the system is run for a prolonged duration because a small coil such as this does not have much surface area to radiate heat.

 

The final Tesla resonator has the following specification:

Diameter:

46 

mm

Winding length:

102

mm

Number of turns:

210

turns

Wire guage:

30AWG (0.125mm diameter copper)

 

Wire length:

30.3

m

Inductance:

680

uH

Natural resonant freq:

3.84

MHz

Self capacitance:

2.53

pF

DC resistance:

10.5

Ohms

Base Impedance:

110

Ohms (at 3.84MHz no breakout)

Note the measurement for inductance of 680uH. That is 0.68mH! A figure that would be considered very low for a normal Tesla Coil resonator. But at least there is one advantage to winding resonators for 4MHz. When compared to more traditional frequencies they are small and don't have many turns so they are quick and easy to wind by hand !

The resonator is used with no Toroid as I have not found large toroids to have a beneficial effect on CW coil designs. However a pointed breakout point is connected to the wire leaving the top of the resonator and protrudes well into the air above the resonator. This provides a point for the corona to form and prevents sparks jumping off from the top turns of the resonator. Such breakout from the wire itself results in rapid burning down of the wire. Therefore a metal bolt or screw is employed for the breakout point, and is replaced when the point burns down.

Initial tests on the resonator showed that around 50 watts of power were required to achieve breakout. Beyond that, any further increase in power would increase the size of the corona formation at the breakout point on top of the resonator. It was estimated that around 300 to 500 watts should be sufficient to produce a reasonable display of corona, without burning up the relatively small resonator. So it is now merely a task in generating the necessary RF power at 4 Megahertz.

 

Producing the necessary RF power

Conventional power amplifiers for use at several Megahertz usually employ specially designed RF power semiconductors and operate in Class A or Class C. These devices typically use special fabrication techniques that reduce the stray capacitances down to acceptable levels to permit efficient operation in the MHz. These are expensive devices due to the relatively low demand when compared to standard power MOSFETs. Standard MOSFETs are manufactured in the millions for switched mode power supplies and motor drives, so the market price is very low. Traditional RF power amplifiers typically have efficiencies in the range of 30% to 70% depending on the design and mode of operation.

For this project I wanted to use general purpose components that are easily available. This would enable anyone else to copy it easily and with good success without having to spend a lot of money on exotic semiconductors. However, there are very real challenges associated with getting standard switch-mode power MOSFETs to switch efficiently at 4MHz. These problems are not without solutions however.

At a frequency of 4MHz the following problems exist with using standard switch-mode power MOSFETs.

1. Difficulty in driving the gate with fast edges due to its large capacitance.
Large die power MOSFETs typically have gate capacitances of several nanofarads. It takes considerable power to quickly charge and discharge this capacitance 4 million times every second! The "Miller effect" adds to this difficulty due to capacitance from drain to gate if the device is operated from a high supply voltage.

2. High switching losses due to the switching time being a significant portion of the total period.
A typical switching time of 50ns is acceptable at 200kHz where the switching time only represents 1% of the total period. However the same 50ns switching time represents 20% of the period at 4MHz. This means that the device would spend a considerable portion of the total period passing through the linear region dissipating a lot of power.

3. High switching losses due to repeated discharging of the MOSFETs output capacitance.
This is the real killer. Every time a MOSFET is switched on the energy stored in its output capacitance is discharged into the MOSFET channel. This is particularly bad for large die devices being operated at high voltage and high frequencies, when the charge stored here can be considerable. For example an IRFP460 device with 320pF of output capacitance has 14uC of energy stored there if the drain goes to 300V when the device is turned off. This may not sound like much energy until you consider that this energy is discharged into the channel every time the device is turned on. This gives a power dissipation of 56 watts in the transistor just due to the devices own output capacitance!

MOSFET device capacitances

So how do we solve these problems ?

 

Enter the Class E amplifier…

The Class E RF power amplifier is a switching circuit that was developed by Nathan and Alan Sokal in 1975. Although it is quite old, it has only recently gained in popularity due to advances in semiconductor technology and better understanding of its advantages. The Class E amplifier makes use of a resonant circuit to get around the usual problems of switching power semiconductors at high frequencies. This topology is now finding commercial success in high power radio transmitters and RF generators for plasma generation and induction heating. High voltage MOSFETs are the usual choice of switching device.

In short it uses the damped ringing of a resonant circuit to force the voltage across the switching device to zero just before it turns on. In practice the voltage across the device is force to zero, and the current through the device is also zero at the instant when the MOSFET is instructed to turn on. This eliminates switching losses at turn-on because there is no voltage or current present, and therefore there is no energy stored in the device's output capacitance. Also the "Miller effect" is minimised since there is no high voltage across the MOSFET at the instant when it is turning on. Therefore the gate drive circuit is able to turn on the device in the shortest time possible.

A simplified circuit for a Class E amplifier is shown below, with an accompanying explanation of its operation. V is the DC supply voltage. S is a high speed switch such as a MOSFET. L is the resonant tank inductor. C is the resonant tank capacitor, and R is the load resistance that also damps the tank circuit. Together the three components L, C and R form a damped parallel resonant circuit.

Charging the Inductor

When switch S is closed, the power supply voltage V appears across inductor L.

This causes current to flow from the supply V through L

Current through L to increase linearly, storing up energy in its magnetic field.

Damped ringing of resonant circuit

When switch S is opened, the energy stored in the inductor is released to the tank capacitor C and the load resistance R.

If S is left open then energy is repeatedly exchanged between L and C and a damped oscillation takes place until all the energy is eventually dissipated in the resistive losses of the load R.

The key to Class E operation is to arrange for the first trough of the damped oscillation to return to zero volts, and close the switch again at this precise instant when the voltage and circulating current are both zero.

Class E relies on a damped ringing circuit

In practice capacitor C is often placed directly across the switch. This is electrically equivalent since at high frequencies both supply rails can be considered to be the same point due to the supply decoupling. When a MOSFET is employed as the switch S there is already some capacitance across S due to the devices own output capacitance. In this case the value of C is adjusted to allow for the ammount already present due to the MOSFET's output capacitance. Connecting C directly across the switch is also desirable so that the current flow at turn-off of the switch is diverted by the minimum distance across the PCB when it switches from flowing through S to flowing through C. This helps to reduce ringing.

An alternative arrangement

In a similar fashion the load resistor R can also be placed across the switch. This can sometimes be more convenient since one end of the load is now connected to ground.

It does not really matter where the load resistance is placed, as long as it provides the necessary damping action on the resonant circuit. So there are many variations of the Class E amplifier circuit, but all rely on forcing the voltage and current seen by the switch to zero at turn-on.

 

In a Class E RF amplifier a tank circuit is used to provide the resonance, and the load impedance provides the necessary damping. The values of the tank circuit inductor, capacitor and the load resistance are all calculated to get the correct frequency and damping. When this is accomplished the drain voltage lands smoothly at zero volts just before the MOSFET is given the instruction to turn on. The resonant frequency of this circuit is deliberately designed to be somewhat lower that the actual operating frequency of the amplifier in order to provide the necessary inductive reactance to the MOSFET for correct operation.

If the load resistance is too high, then Q factor is low. In this case the resonant circuit is too over damped, and the drain voltage does not ring all the way back down to zero. This causes the MOSFET to turn on with some voltage across it and hence some charge is stored in its drain-source capacitance. This causes switching losses at turn on because voltage and current are not zero at this time, and energy from Cds is discharged into the channels "on-resistance".

Ringing is over-damped

If the load resistance is too low, then the Q factor is high. In this case the resonance is under-damped, and the drain voltage attempts to ring below the zero volt rail. However this cannot happen. The body-diode of the MOSFET is forward biased preventing the drain voltage from going negative with respect to the source. This causes current flow through the body diode which must then be transferred back to the channel at turn on. This results in switching losses because of charge that must be swept out of the body-diode when it recovers, and the fact that there is current flowing at the instant when the MOSFET is turned on.

Ringing is under-damped

In an RF amplifier it is necessary to include the resonant tank circuit in the amplifier itself because the load (the antenna) typically has a very low Q factor. It is therefore too highly damped to be able to drive the drain voltage back to zero on its own. So a separate resonant tank circuit is built into the amplifier. However, in our Tesla Coil driver application, the load is the TC resonator which has a fairly high Q even after spark breakout. With considerable power going into corona the damping of the resonator is sufficiently light. This means that the TC resonator itself can provide the damped resonant action required for correct Class E operation. Therefore no additional resonant tank circuit is required as part of the amplifier itself. This is beneficial because it simplifies the design. Inductors and capacitors for high power operation at 4MHz also tend to be bulky, expensive and lossy too.

It is worth bearing in mind that as the power level is increased, the corona structure at the top of the resonator grows and the damping increases. It is anticipated that beyond a certain power level the Q factor will fall below that required to achieve correct Class E operation. In short there will be insufficient circulating current to drive the drain voltage all the way back down to zero volts before the MOSFET turns on. Beyond this power level the benefits of class E operation are lost, efficiency falls, and power disipated as heat in the MOSFET will increase rapidly. Inclusion of a resonant tank circuit in the Class E output stage might increase the Q factor and regain correct operation but it is not anticipated that this would be required at the power levels encountered in this small system.

It is also worth noting that detuning of the resonator or arcing to an object also disturb the system from its ideal Class E operating point. Detuning by nearby objects or arcing to ground all decrease the operating efficiency of the amplifier and increase power dissipated as heat in the switching MOSFET. This should be kept in mind to avoid damage to the MOSFET when demonstrating the system under a variety of sparking conditions.

Advantages of the Class E power amplifier:

1. Simple design uses only one switching device.
2. Eliminates turn-on losses due to stored energy in the devices output capacitance.
3. Eliminates turn-on losses due to overlap of voltage and current when device turns on.
4. Zero-voltage at turn-on eliminates Miller effect due to drain-gate capacitance.
5. Minimises turn-off losses by limiting rate of rise of voltage when device turns on.
6. Circuit makes use of stray capacitance and inductance that would normally be hindrance.
7. Very high efficiency (90 - 95%) so very little power to be dissipated as heat.
8. Low EMI and harmonic output because of the smooth drain voltage and soft switching.

Perhaps the main disadvantage of the Class E arrangement is that it is relatively intolerant of incorrect matching into its load. Whilst load matching is usually good a radio transmitter, this intolerance to mismatching may cause problems when put to service as a Tesla Coil driver.

 

The HF-SSTC circuit explained

In many ways the HF-SSTC is like a high power radio transmitter. It consists of an oscillator followed by a series of amplifiers to reach the desired power level at the output.

The full schematic can be seen by clicking here.

Alternatively, the system diagram can be broken down into the following functional blocks:

1. 16MHz crystal oscillator.
2. CMOS divider stage to produce 4MHz.
3. Bipolar push-pull stage to drive first MOSFET.
4. IRF520 Class E driver stage.
5. MTW14N40 Class E high power amplifier
6. Matching into the Tesla resonator

HF-SSTC block diagram

Each of these blocks will be explained in detail below…

1. 16Mhz crystal oscillator module.
A crystal oscillator module running from 5 volts provides a highly stable square-wave at 16Mhz. It was decided to run the circuit from a fixed frequency oscillator for several reasons. Firstly the bandwidth for correct operation of Class E amplifiers is quite small. Secondly, it is anticipated that such a small Tesla Coil will not produce enough corona even at full power to greatly detune the resonator. A small system should also be less sensitive to the surrounding environment detuning the coil. Therefore a fixed frequency source was chosen. The system can always be tuned by adding or removing a turn from the Tesla Resonator, or fine adjustments made by raising or lowering the breakout point.

2. CMOS divider to produce 4MHz.
This stage consists of a CD4024B ripple counter running from a 12 volt supply. It serves several purposes in this design. Firstly it divides the 16MHz crystal signal by four to obtain a 4Mhz signal that matches the resonant frequency of the resonator to be driven. Secondly it translates the level from the 5 volt crystal module up to the 12 volt supply rail, giving a squarewave output that is 12 volts peak to peak. And thirdly the division process ensures that the resulting drive signal at 4MHz has equal mark and space durations. I.e. its duty ratio is equal to 0.5.

It is believed that running the oscillator at a higher frequency than the output power frequency results in good immunity to feedback. This can be contrasted with a system in which a low power 4MHz crystal oscillator would be very susceptible to any feedback caused by the strong E-field of the nearby Tesla Resonator. Therefore the additional complication of the frequency divider stage is thought to be worthwhile to increase immunity to interference from the power stages.

3. Biploar push-pull stage.
This consists of a pair of complementary ZTX450 and ZTX550 transistors. These transistors are configured as a push-pull emitter follower pair and therefore provide current gain. Their job is to provide sufficient current at 4MHz to properly drive the IRF520 MOSFET in the first RF amplifier stage. This requires approximately 200mW of power, which can not be supplied by the CD2024B device alone.

The output of the ZTX450/ZTX550 stage is a squarewave at 4MHz that is almost 12 volts in amplitude, but has considerable current sourcing ability. The 4MHz squarewave from the push-pull bipolar stage is coupled into the gate of the following IRF520 MOSFET in the driver stage. This is done via a matching network consisting of a series resonating inductance and a ferrite transformer. This effectively matches the gate-loss resistance of the IRF520 device to the output of the bipolar stage so that the MOSFET gate is driven efficiently with a strong gate signal at 4MHz. In short, the ferrite transformer transforms the very low gate loss resistance up to a value of a few tens, and the series inductance cancels out the capacitive reactance of the MOSFETs gate at 4MHz. The gate of the MOSFET then appears like a pure resistance at 4MHz which is much easier to drive.

4. IRF520 Class E driver stage.
This stage makes use of an International Rectifier IRF520 MOSFET running from a 12 volt supply. It is used to develop sufficient power at 4MHz to properly drive the MOSFET in the high power output stage that follows. The MTW14N50 MOSFET in the final amplifier requires approximately 5 watts of power to drive its gate fully for efficient switching. The driver stage incorporates the IRF520 device into a Class E amplifier arrangement so that it can provide the required power efficiently whilst dissipating very little itself power as heat.

The output of the IRF520 device is coupled to the gate MTW14N50 MOSFET via a matching circuit and ferrite transformer. This arrangement serves several purposes. Firstly, it transforms the gate impedance of the MTW14N50 device up to a level that represents the correct load for the IRF520 setting the drive power level. Secondly, series inductance is used to cancel out the large capacitive reactance of the MTW14N50's gate so that it appears like a pure resistive load which is much easier to drive at 4MHz. And finally, it shapes the drain voltage waveform of the IRF520 so that it lands smoothly at zero volts for elimination of switching loses at turn on. The resulting power dissipation in this driver stage is so low that the IRF520 device only requires a small clip-on heatsink.

5. MTW14N40 Class E high power amplifier.
The On-Semiconductor (Motorola) MTW15N50 device in the final power amplifier stage runs from 150 volts and switches at 4MHz. It is this device that produces the hundreds of watts of RF required to drive the Tesla Coil and make sparks. The MTW15N50 device is incorporated into a Class E amplifier arrangement and drives the TC resonator via a link-coupled primary winding. The link coupled primary reflects the damped resonant nature of the Tesla coil back to the drain of the MOSFET. The number of primary turns, coupling coefficient and tuning of the secondary are all chosen in order to obtain the correct voltage profile across the MOSFET for high efficiency Class E operation, whilst at the same time ensuring that power is coupled into the resonator.

This means that the device turns on with zero-voltage across it and zero load current at that time. This zero-voltage zero-current switching means that the primary sources of power dissipation are conduction losses due to the drain current, and dissipation of the gate-drive power in the MOSFET gate structure. Although the efficiency of this final stage is greater than 90%, this still implies that around 40 watts or more must be dissipated in the switching MOSFET. Therefore this device requires a considerable heatsink to dissipate this power plus the gate drive power if continuous operation is desired.

6. Matching into Tesla Resonator.
This is where the RF power developed by the previous stage is coupled into the Tesla Resonator. In this circuit the matching arrangement appears deceptively simple, being merely a few turns of wire around the base of the resonator. The dimensions, number of turns, and coupling coefficient to the resonator are critical to correct operation however. Both the coupling and the inductance of this primary winding determine the power throughput of the system. They also effect the loading (or damping) on the final Class E power amplifier stage. And must be adjusted just right to get the drain voltage waveform to land smoothly at zero volts at the instant before the MOSFET is turned on.

The value of the coupling coefficient also effects the tuning of the resonator. The reason for this may not be immediately apparent. But, increasing the coupling coefficient ties up more of the lower turns of the resonator in transformer action by magnetic coupling to the primary. As the coupling coefficient is increased there is less of the resonator's total inductance free to resonate, and so the apparent resonant frequency increases.

Due to all of these interactions, the choice of primary turns and coupling coefficient is roughly 50% calculation, and 50% trail and error. In practice I was lucky and quickly arrived at a combination that gave good spark output and the correct Drain voltage waveform for efficient operation of the MOSFET. It is then simply a matter of making small alterations to the number of turns, raising or lowering to alter coupling (and tuning) and altering the value of the shunt capacitor in the final stage to increase or decrease the power level.

Less primary turns, increased coupling and a larger shunt capacitor across the MOSFET all increase the power throughput and make the system draw more current for a given supply voltage. Conversely, more primary turns, decreased coupling and a reduction in the shunt capacitance all decrease the power throughput and current consumption.

The most important goal is always to achieve zero voltage turn-on of the MOSFET since this stops it self-destructing due to excessive switching losses. Once this is achieved with a low supply voltage, good spark output is guaranteed at higher supply voltage with minimal adjustments.

 

Actual waveforms

This shows that the voltage observed at the MOSFET gate is approximately sinusoidal. The flat tops are due to harmonic content coupled back from the output side of the Class E amplifier. The "cusps" on the rising edges of the gate waveform are attributed to the Miller Effect limiting gate dv/dt as the device turns on.

This shows the classic Class E voltage profile observed across the MOSFET switch. At turn-off the voltage rings up to over 400 volts and then lands softly at 0 volts before the device turns on again. The slight blurring of the trace is due to the varying load presented by the corona discharge.

 

Spark appearance

HF-SSTC running 300 watts at 4MHz.

Corona is silent flickering orangey-lillac plasma flame.

Arc to a grounded wire.

A silent pale white arc to ground.

 

Future developments

I am fairly happy with the status of the HF-SSTC as a miniature demonstration coil so I don't really have any grand plans for improving it. I am just glad that it actually worked and proved that a solid state TC at 4MHz can be done with standard parts. However, here are a few things that could be done to further develop the project...

 

Pulsed operation.

The HF-SSTC circuit was originally designed for CW operation but it can easily be modified for pulsed operation using an interrupter. This can be done by either gating the drive to the final Class E amplifier, or by gating the +150V power supply to the final amplifier. Both of these techniques are very effective but the former is easier to achieve by simply applying the gating signal to the reset pin of the CD2024B divider IC. A NE555 timer is ideally suited to this provided that it is well screened from the high E-field present around the resonator.

If gating the power supply to the IRF520 driver stage, care must be taken to ensure that it rises and falls quickly. If this is not the case, the MOSFET in the final power stage may see insufficient drive to turn it on completely for some time at the start and end of the burst. This causes massive power dissipation and ultimately destruction of the MOSFET.

Some experiments were done feeding the interrupter signal into the reset pin of the CD2024B and this was very successful. Tests using 200us bursts at 200Hz repetition rate yielded a display of delicate branched sparks and loud bussing sound, in place of the previous silent plasma flame. Average current consumption was reduced considerably and the MTW14N50 device in the final power amplifier remained cold.

#### Add picture here ####

 

Audio modulation (singing arcs)

The stable nature of the corona at these high frequencies make the HF-SSTC an ideal candidate for audio modulation. Audio modulation of any Solid State Tesla Coil is possible by varying the RF output power in sympathy with an incoming audio signal. This varying power causes the air around the spark to vary in temperature and expand and contract producing sound waved. Sound produced in this way is fairly high quality because there is no inertia and there are no resonances caused by moving parts. But there is usually a large amount of background hiss due to the corona itself. This tends to detract from the quality of the music. However, by increasing the operating frequency up into the Megahertz range the corona becomes much more stable and the hissing disappears. This allows for high quality audio modulation with minimal distortion or background noise.

Audio modulation of this system is best implemented using high level modulation of the +150 volt supply to the final amplifier stage. This provides minimal distortion and ensures that the Class E amplifier always operates near to its intended operating point for good efficiency. This is identical to the way in which modern high power AM radio transmitters work. Frequency modulation techniques are not recommended for implementing audio modulation with this design due to the relatively narrow bandwidth of the Class E amplifier for efficient operation.

 

Self oscillating design (frequency tracking)

The HF-SSTC system is small and only operates at a few hundred watts so the resonator does not seem to detune by a very large amount when producing corona. However, one improvement to the design would be to make the system track the resonant frequency of the Tesla Resonator with changes in corona loading or objects being moved nearby the resonator. There are a variety of ways that this can be implemented including various feedback schemes involving antennas, Phase Locked Loops or sensing the base current of the TC resonator.

Sensing the E-field around the secondary and feeding this back to the input is by far the simplest way to achieve a self oscillating design that will track the resonant frequency over a small range. The whole HF-SSTC system will oscillate if some of the output signal is fed back to the input and the right conditions exist. In order to sustain oscillations the gain around the loop must be greater than one, and the total phase shift around the loop must equal 360 degrees.

In practice the E-field around the resonator is sensed with a small antenna nearby. The signal is then "squared-up" with either a comparator or CMOS inverter IC. This signal is then fed back into the ZTX450/ZTX550 push-pull stage that drives the Class E power amplifiers. The signal is either inverter or not inverted to obtain the necessary total phase shift around the loop to sustain oscillations.

Whilst this arrangement is simple, functional, and is certainly superior to fixed frequency operation, it does have some drawbacks. Firstly, it requires the antenna to be touched in order to provide a small pulse to start the system into oscillation. It also has limits on the range over which the frequency will be tracked before problems are encountered. (This is mostly due to the narrow bandwidth behaviour of the Class E amplifier stages.) A more elaborate system using a Phase Locked Loop (PLL) can offer advantages in the area of start-up and give better control over the tracking range.

 

More power ?

The output power of the final MTW14N50 class E power amplifier is limited by the devices RMS current rating and its avalanche voltage rating. The output power is also indirectly limited by the heatsink's ability to get rid of heat generated by the device.

The MTW14N50 device is rated for 9 Amps of drain current at 100'C and has a maximum drain voltage rating of 500 volts. In a class E amplifier the peak drain voltage is equal to approximately 3.3 times the supply voltage when tuned correctly. Similarly, the RMS drain current in this arrangement is approximately 1.8 times the supply current. These limit the power supply to a maximum of 150 volts and 5 amps respectively. That equates to a maximum input power of 750 watts. With a typical efficiency of 90% that would deliver around 675 watts of power at 4MHz to the Tesla resonator. The remaining 75 watts of power would have to be dissipated as heat in the MOSFET device. So this would require a large heatsink to keep the junction temperature within safe operating limits.

It should also be noted that these calculations allow no margin for mis-tuning of the system if something goes wrong. Under these conditions the drain voltage and drain current can increase and switching losses can soar if the zero-voltage switching condition is lost. Therefore it is always wise to build in some safety margin for mishaps. This is especially true for a Solid State Tesla Coil where the load is likely to vary as objects are brought near to the resonator, or things arc where they were not supposed to!

In practice an output power of 500 watts was considered to provide a good compromise between spark output and some tolerance to mismatch.

 

How to get more power ?

There are a number of ways that we can produce more power at 4MHz. These include:

 

1. Choosing a larger switching device.

A larger die power MOSFET could be employed to increase the current carrying ability of the final stage, or a device with a higher breakdown voltage could be chosen to allow the supply voltage to be increased. Both of these would allow more output power to be generated at 4MHz, but it must be kept in mind that larger die MOSFETs have larger input and output capacitances. This means that a larger device will require significantly more drive power to fully drive the gate, and the shunt capacitance from drain to source will also need to be adjusted to allow for the difference in the devices own output capacitances.

Minority carrier devices such as IGBTs are not considered suitable for use at this high frequency due to there comparatively long "current tailing" period after turn-off.

 

2. Paralleling Class E amplifiers.

It is generally not good practice to parallel MOSFETs directly at their terminals, however several Class E amplifier stages may be connected in parallel. The important thing to remember is that the amplifiers are tied in parallel at high impedance points such as before the gate drive transformer, and after the output matching network if one is employed. Paralleling complete amplifiers in this way helps to ensure that the load is shared evenly between the MOSFETS.

However, it is very important to make certain that the devices and layout of the amplifiers being paralleled is identical to ensure that they operate in phase and share load evenly.

 

3. Double-ended Class E drive.

Two Class E amplifiers can be employed in a double ended arrangement to obtain more output power. In this arrangement the two amplifiers are fed with complementary drive signals (180 degrees out of phase) and the TC primary winding is connected between the drains of the two MOSFETs. This doubles the voltage applied across the TC primary winding and gives a considerable increase in output power without requiring larger switching devices. It also does not require such care in the matching of the two amplifiers that is so important in parallel operation.

 

4. Class DE half bridge (or full bridge)

4MHz is around the limit of frequency at which a half-bridge or full-bridge of standard MOSFETs can be made to operate with reasonable efficiency. This is done by using the devices output capacitances alone to cause the mid-point voltage to slew smoothly from one supply rail to the other before each MOSFET is turned on. This gives zero-voltage turn-on and eliminates energy stored in the device's output capacitance in a similar way to our Class E amplifier described previously.

This Class DE inverter works similar to a conventional half-bridge or full-bridge, but the operating frequency, output tuning, switching deadtime and layout are all critical to achieving zero-voltage switching at such a high frequency. It can be made to work and output several kilowatts of power in the low Megahertz. However, if the zero-voltage zero-current switching condition is lost for any reason the inverter will self-destruct because of excessive switching losses, shoot-through, or because the body-diodes of the devices are too slow for switching at these frequencies.

See these papers for further details: 

High Power & High Frequency Class DE Inverters

#### Add link to Directed Energy Class DE paper ####

 

5. Use of an additional power amplifier.

A commercial RF power amplifier intended for HF communications could be employed to obtain greater output powers. The designer would only have to design a matching network to match the 50 ohm output of the power amplifier into the Tesla Coil resonator with breakout. Initial matching would typically be done by an educated guess, then refined to minimise the reflected power indicated on the RF power amplifier.

 

  Back to home page