|
|
|
|
Introduction
As a power electronics engineer, I frequently work with large semiconductors in power supplies and motor drives, etc. These often switch thousands of watts at several hundreds of kilohertz. Modern power transistors offer an increasingly viable alternative to the Vacuum Tube Tesla Coils, as performance improves and prices continue to fall.
Whilst testing a switch mode power supply for a customer, the TC resonator at the end of the bench caught my eye, and curiosity got the better of me. I could not resist the temptation to see what would happen if I replaced the high frequency transformer in the supply with a primary coil feeding the resonator. The worst thing that could happen was that the power transistors would fail catastrophically , and after all the supply wasn't mine anyway ;-)
It actually worked surprisingly well (for a few seconds), and I decided to design my own solid-state mini coil.
The first design was based around two IRF740 MOSFET devices made by International Rectifier. The two switching devices are connected in a half bridge configuration as shown in the schematic below. These devices are very close to the theoretical "ideal switch". They can switch 400volts at 10amps in around 50 nanoseconds and are reasonably priced.
The half bridge is fed from the 240Vrms mains supply, and the MOSFET devices are turned on alternately at roughly 250kHz. The high voltage square-wave output from the transistors is fed into a 25 turn primary which is tightly coupled to the bottom portion of the resonator. At resonance the base current of the resonator is sinusoidal, and a sinusoidal current flows in the primary coil also.
|
Simplified circuit: |
Primary voltage waveform (Red square wave,) |
If the Tesla Coil is driven at its resonant frequency then the switching transitions of S1 and S2 occur when the current (Ip) passes through zero. This means that switching losses in the MOSFETs are practically eliminated, and heating is due to conduction losses only. (This is technique is explained as "soft-switching" or "ZCS" in many Power Electronics papers.)
|
|
An advantage of the primary feed method is that it provides the necessary voltage transformation required to match the output impedance of the inverter to the resonator. This negates the need to employ a separate high frequency matching transformer or the use of elevated supply rails to get the required drive voltage. A significant disadvantage of the primary feed method is that very tight coupling is required (k>0.35) in order to get good power transfer. This makes insulating the primary from the secondary somewhat challenging as the power level is increased. |
The drive electronics is based around the TL494 PWM controller IC made by Texas Instruments. This IC is fairly "long in the tooth" but it is well behaved and is also easy to obtain. The IC contains an internal sawtooth generator and the necessary comparators and latches to produce the drive signals required for each MOSFET in the half bridge. The IC generates two complementary drive signals with a short dead time between transitions to ensure that one MOSFET has had time to turn off before the opposing device is turned on. Without this precaution the conduction times of both devices can overlap shorting the mains supply with interesting (read expensive,) consequences.
|
Click here to view
|
The two outputs from the TL494 are boosted in current by push-pull stages and are used to drive the primary of a small ferrite transformer. This transformer serves to isolate the sensitive low voltage control circuitry from the high power MOSFET side, whilst coupling the drive signals to the gates of the two MOSFETs. (Power semiconductors usually fail short-circuit. Without this isolating transformer such a failure would almost certainly lead to damage of the control circuitry too.) The latest schematic can be found further down this page. |
This isolation transformer has two secondary windings wound in opposite directions to drive the gates of each MOSFET. This serves two functions. Firstly, it ensures that when one MOSFET is turned on by a positive gate voltage, the opposing device is held firmly in the off state by a negative gate voltage. (This negative bias is useful to prevent spurious turn-on due to the Miller capacitance from drain to gate of the MOSFET.) Secondly, the two isolated secondary windings allow the high-side (top) MOSFET to be driven without the need for complicated floating or bootstrap power supplies.
The overall arrangement is very flexible because the oscillator is continuously running. The MOSFETs merely chop up the supply voltage as instructed by the oscillator and feed the RF to the primary coil. This means that the supply voltage to the half bridge can be DC or virtually any desired waveform you choose to throw at it.
The effect of varying the supply voltage is to Amplitude Modulate the RF applied to the TC primary winding.
I tried 4 different supply schemes which gave different RF envelopes and radically different spark characteristics:
Half wave rectification,
|
RF envelope: |
This was achieved by inserting a diode in series with the mains supply to the MOSFET half bridge so that only positive half-cycles resulted in current flow. (This is necessary anyway in order to prevent shorting of the negative supply cycles by the MOSFET body diodes !) The RF envelope consisted of rounded bursts of RF lasting 10ms with 10ms gaps in between. |
|
Spark appearance: |
Sparks were roughly 6 inches long, very straight and "sword-like" in character. The absence of branching in the streamers struck me as being very odd. Apparently this appearance is common in Vacuum Tube TCs also. The sound was like a muffled 50Hz buzz but still quite loud. Power is estimated to be around 160 watts in the picture opposite. |
Full wave rectification,
|
RF Envelope: |
This was achieved by using a full wave bridge rectifier between the mains line and the MOSFET bridge. This ensures that there is current flowing through the inverter during the entire supply cycle. The power drawn from the mains line roughly doubled as expected and the RF envelope assumed the classic full-wave rectified shape. This implies a considerable increase in the average RF energy applied to the TC. |
|
Spark appearance: |
Sparks became noticeably fatter and more bushy, but there was no increase in length. The picture opposite clearly shows the greater "fullness" of the discharge including wispy branches leading off from the main feature. The tone of the sound changed to twice the pitch (100Hz) and became distinctly more "full-throated" and hissy. Power is estimated to be around 300 watts. |
|
|
|
Smoothed DC,
|
RF Envelope: |
This was achieved by using a full wave bridge rectifier and a large high voltage reservoir capacitor ahead of the MOSFET bridge. This provides a constant supply of around 350VDC to the inverter. Power draw increased again due to the sustained high voltage, and the RF envelope was that of a continuous-wave source. During this test some warming of the MOSFET heatsink was noted due to the high average current. |
|
|
The discharge from the breakout point became very bushy. It looked and sounded like a jet of burning gas, and spread out like a cone from the discharge point. All of the buzzing was gone to leave a pure hissing sound. This test produced a lot of ozone really quickly and also overheated the thin wire at the base of the secondary coil blistering the varnish. Power is estimated at 420W in the picture shown opposite.
|
Phase angle controller,
|
RF Envelope: |
A phase angle controller (similar to a commercial light dimmer) was connected ahead of the MOSFET inverter in order to interrupt the supply to the half-bridge. The phase angle controller was set to turn on exactly at the peak of the mains supply cycles and remain on until the end of each half-cycle. This leads to a very sharp rise in the voltage applied to the inverter, rising from 0 to around 350 volts in a matter of microseconds. This sudden application of power results in a sharp rise in the RF envelope and an interesting effect on the spark characteristic. |
|
Spark appearance: |
The discharge from the breakout point became branched like a conventional spark gap TC. The sparks were about 6 inches long, distinctly spidery and danced about frantically. The sound was considerably sharper and more raspy, no doubt due to the rapid rise of the RF envelope. It was similar to the sound from a conventional 100BPS synchronous TC, but sounded slightly deeper and more fuller. RMS Power level was thought to be around the 180 watt mark, although this measurement may not be particularly accurate. |
|
|
Another picture showing a peculiar branching in the discharge. The arc to the lower right of the picture is striking a piece of metal which was not earthed. |
Unlike a conventional damped-wave Tesla Coil, the solid state Tesla coil is capable of producing considerable amounts of sustained RF power. This leads to a few unusual things:
Firstly, the base of the secondary became very hot due to the high RMS current flowing through the fine wire. Maybe skin effect plays some part in this also. This is particularly noticeable if the system is run in CW mode for any length of time.
|
|
|
There is visibly more current in ground strikes than found with my spark gap TC. Sparks to ground appear like pale ghostly white flames and arch upwards with the heat like the arc from a Jacobs Ladder. Anything flammable catches fire instantly in the arc. |
I noticed that I got tiny RF burns if I touched anything metallic in the vicinity of the running coil even at fairly low power levels.
At one time I forgot to put the breakout point on the solid state coil, and an unused resonator about 2 feet from the solid state coil, (but quite close to me,) sprang to life with a firey crown of corona. Boy did that surprise me !!!
Please note that building a solid state tesla coil IS NOT EASY. In fact both the design and construction present significantly different and far more complex challenges than those encountered in conventional tesla coil work. A very fast (100MHz) oscilloscope and a large bag of MOSFETs are essential. The biggest problem with this type of design is that a blown MOSFET is often the first sign you get about an underlying problem, so diagnosing the cause of blown semiconductors can sometimes be difficult. Despite these difficulties, the Solid State coil is a beautiful thing when it works properly.
I think that a small solid state (or Vacuum tube) Tesla Coil is the method of choice for "up-close" analysis of spark behaviour and demonstrations. When compared to a conventional coil it has the following characteristics:-
Less audible noise. Not so much crash and bang, more hum and hiss.
Less RF hash radiated. The SSTC is very clean due to its continuous electronic source of RF, and causes no TVI.
Higher average RF power. The SSTC seems to produce a much stronger RF field than a similarly rated spark gap TC.
Great for corona displays and lighting neon tubes at a distance without wires.
Wide variety of spark characteristics from forked lightning to flames by modulating the RF generator in different ways.
Does not shock so much as burn, but such RF burns are reported to be very nasty.
Ideal for research because the system is under full electronic control. (Maybe one could adjust the burst rate fast enough to play the national anthem ?)
Significant downsides to the solid state approach are as follows:-
Requires that the designer have a good working knowledge of power electronics,
Careful attention must be given to layout, and screening,
Suitable semiconductors are moderately expensive,
Semiconductors are still quite fragile in such an application, and are not forgiving of any mistakes.
Application notes and design examples provided by device manufacturers help greatly with design and layout tips, and power semiconductors are consistently becoming faster, more robust, and cheaper, so the future looks promising for the Solid State Tesla Coil.
Late last year I tackled the reliability problems with the original SSTC design. I also modified the circuit to form a full H-bridge configuration in search of some longer sparks.
|
|
The resonator pictured here is 3.5" x 16" and is topped with a 6"x1.5" toroid. Sparks look similar to those from a conventional spark gap Tesla Coil, although they are somewhat thicker and hotter. The driver runs directly from the 240V 50Hz AC mains which is then half wave rectified. Current draw is approximately 5 amps RMS. It uses four STW15NB50 MOSFET devices connected in a H-bridge arrangement. This drives the two ends of the primary coil in opposition (anti-phase) and effectively doubles the voltage swing that can be developed across the primary winding. (This really helped achieve a good spark length.) |
|
There are currently no smoothing or energy storage capacitors in use here. The primary consists of 19 turns of wire and is link coupled over the bottom third of the resonator. (k estimated at around 0.40) The RF peak envelope power has been measured at 4800 watts, so the RMS power input should be about 1200 watts or so. Peak RF current in the primary is 22 amps, and the resonator appears like a constant current sink once the breakout potential has been reached. The resonant frequency is nominally 350kHz, but the frequency of the driver is dynamically swept throughout the mains supply cycle in an attempt to maintain correct tuning as the sparks grow. This dynamic tuning is of some importance to achieving long sparks. Without some automatic adjustment of the oscillator the growing sparks "snub" themselves out as they detune the resonator and limit the terminal voltage. Dynamic tuning is achieved by feeding a small portion of the supply voltage into the frequency determining part of the driver circuit. This causes a progressive drop in the drive frequency as the supply voltage increases and the sparks propagate. It is very crude but definitely makes an improvement. The four MOSFETs are only slightly warm after a 3 minute run, and I have run it for 30 minutes continuously to check reliability. After the longer run, the heatsink was quite warm, and both primary and secondary displayed noticeable heating. Average spark length is around 14 inches with occasional hits out at 18" or so. The sparks generate a loud thudding humming sound, and appear like inch thick flames where they contact the toroid. I have also seen several brilliant white balls emitted from the toroid during operation. (See frame sequence opposite.) These are thought to be balls of burning Aluminium which come from the surface of the foil covered toroid, although it really surprised me when it first happened ! The surface of the toroid is covered with small 1/8th inch foil bumps to promote breakout. If a smooth toroid is fitted without any breakout points, there are severe flashovers which instantly burn up the plastic primary form: A significant problem with using such a tight coupling. I have no plans to increase the spark length of this particular design, as it is intended as a compact portable unit. It will be packaged neatly and used for demonstrations at Teslathons etc. However, I may try to build a bigger solid state system in the future, as the flexibility and absence of TV and radio interference is appealing to me. Several members of the Tesla List have also pointed out that this coil represents a good platform for investigating the little understood areas of spark loading and impedance matching to corona. |
|
|
|
The picture opposite shows a fierce 9 inch flame produced by running the Solid State driver from a continuous smoothed DC supply. (CW mode) This causes noticeable heating in the driver, the primary winding, and the lower portion of the Tesla resonator. Input power was measured at 1500 watts. Despite this low power the end of the breakout point (small terminal driver) became melted into a ball. The tip continued to glow for seconds after the power was switched off. The discharge was very hot in this mode, and "heat-shimmer" was visible above the corona. Sound was a rushing, hissing, crackling noise. |
|
|
|
An unusual phenomena frequently observed above power arcs. In these two frames the coil is arcing over a distance of 6 inches to a grounded wire. Brief smudges of pale yellow light are seen rising above the main flame-like arc. I have been informed that this is due to combustion of trace gases in the air. |
The latest schematics for the control electronics and power electronics can be downloaded by clicking the two links below. These schematics have been checked for obvious mistakes, and are believed to be error free. Significant improvements have been made from the original design in the following areas:
Re-configuration of the MOSFET gate drivers to reduce the dead-time at switching transitions. The old design had a large 5% dead-time at switching instants to allow one MOSFET to turn off before the other is turned on. This dead-time was found to be excessive, and caused the body diodes of the MOSFETs to conduct heavily due to the free-wheeling current.
Isolation of the MOSFET body diodes, using series a Schottky diode and parallel fast recovery diodes. This eliminates problems due to the slow reverse recovery characteristics of the body diode. This modification combined with the one explained above, dramatically reduces MOSFET mortality rates !
Dynamic tuning. The carrier signal generated by the TL494 is "Frequency modulated" by the HV supply voltage. The frequency is actually swept down by a few percent as the voltage increases in an attempt to track the resonator frequency as the sparks grow. This is quite rough, as is does not take into account that detuning only happens above breakout, etc. However it has been found to be very effective, most likely because the loaded resonator Q is low, and the tuning range is actually quite broad during sparking conditions ?
|
Click here to view the schematic for the control electronics:
|
. |
Click here to view the schematic for the power electronics:
|
|
|
The picture opposite shows my Solid State Driver board connected to the Tesla Coil primary winding. The PCB measures 6" x 4" and is mounted directly on top of a large Aluminium heatsink for cooling the power semiconductors. The small transformer at the top right of the PCB provides a 15 volt supply to power the control electronics. Everything else runs directly off the 240V mains supply. |
As a warning to anyone contemplating building a similar system, the development of this project was VERY expensive, I spent the equivalent of around 600 dollars on various semiconductors and had to borrow some sophisticated test gear to debug the design. However, now that it works well I think it is beautiful ! Careful attention must be paid to layout, wiring lengths, heatsinking and shielding to ensure reliable operation. Some design and construction tips can be found by clicking the link below.
Click here to see new pictures of a steel breakout point burning away, and music coming from a spark !!!
If you have bothered to read this far you probably want to know more about SSTC operation. But be warned, this is where it gets a bit more heavy...
Click here to read my in depth SOLID STATE DRIVER THEORY pages. (Recommended reading if you are going to build your own driver.)
or click on this link to see some reasons why MOSFET devices fail in solid state TC duty, if you have built your own driver but have problems!
Photo from Cambridge 2001 by Mark Hales.
Here are a few thoughts for future developments on the SSTC theme:
It is planned to modify this existing rig to operate from smoothed DC later this year, for the purpose of investigating corona impedance and loading issues.
More investigation into decreasing the un-loaded resonator base impedance, in order to get more power into the resonator without requiring very a high coupling coefficient. (Possibly winding a physically larger secondary to be operated above a ground plane.)
Dynamic tuning based on sensing of the resonator base current. Essentially the resonator is made to be the frequency determining part of the oscillator, so the driver frequency "perfectly" tracks the resonant frequency during streamer growth. This will also ensure that switching transitions occur at zero current, resulting in reducing switching losses.
Development of a twin SSTC. This should provide longer sparks between towers, without increasing voltage stresses across each tower.
Improve the heatsinking of the present design, as there are still some thermal issues when run times are long.
Many thanks to John Freau, Alan Sharp and Paul Nicholson for providing information, tips and suggestions for my CW coil work.
Here is a link to Alan Sharp's Web Page which contains excellent information about Solid State Tesla Coil design and construction. International Rectifier's web site also contains many valuable application notes covering topics such as MOSFETs drive circuits, etc. Here is a link to John Freau's Web page which contains some information about Vacuum Tube Tesla coils. (Vacuum tube coils are very similar to solid state coils in their method of operation.)
Also be sure to check out the dedicated section of SSTC related links on my main links page. Countless people have contributed ideas to my SSTC work, and many others have built solid state Tesla coils based on information presented here. There is an ever growing amount of information available about SSTC stuff on these sites.
Back to home page