Wednesday, February 3, 2016

The oneTesla TS: Not For the Faint-Hearted


Summary:  This review has important information for anyone who is thinking about building the oneTeslaTS Tesla coil kit.  It can be done, but it is a very challenging project.  Helpful work-arounds for difficult steps in the project assembly are described, and omissions or confusing parts of the manual are explained.

When a company markets a product kit for the "experienced" do-it-yourselfer, you can expect a few bumps along the road.  After all, meeting challenges and overcoming them is part of what doing it yourself is all about.  The oneTeslaTS (the big kit—not the tinyTesla, which is a separate product) promises to make sparks up to 20 inches (50 cm) long.  At a cost of $400, this is a serious commitment, but it is much less expensive than other commercially built Tesla coils, which cost in the multi-kilobuck range.  Partly for research purposes, and partly out of curiosity to see how the oneTesla folks had pulled this off, this reviewer ordered the kit (plus a set of spare parts), and it arrived a few days after Christmas.

First, a brief summary of my qualifications to critique a design like this.  I teach electrical engineering at Texas State University, San Marcos, Texas.  I have a BS from CalTech, an M. Eng. from Cornell, and a Ph. D. from the University of Texas.  I built my first Tesla coil when I was about 12 (with a college-age neighbor) back in the 1960s.  About a decade ago, a colleague of mine teamed with me on building a second Tesla coil from scratch, a unit I still use in research.  Last year I published a textbook, Analog and Mixed-Signal Electronics (Wiley, 2015).  So I think I am qualified to render an opinion on the quality of the circuit design, the manual, and the experience of building the kit.

To cut to the chase:  My oneTeslaTS works, but I ended up having to modify certain steps in the assembly and do some rewiring to get it to work at all.  And my particular unit had a recurrent problem that was remedied only by replacing the main board.  First, a word about the good features of the design.

Good News:  Small and Straightforward

The oneTeslaTS is specifically a double-resonant solid-state Tesla coil (DRSSTC).  Solid-state simply means that the high-frequency AC power that is transformed up by the transformer at the heart of the unit, is produced by transistors instead of by old-fashioned (noisy, inefficient) spark gaps, which is what my first 1960s Tesla coil used, and is what some of the true monsters of the Tesla-coil world still use.  Double-resonant means that there are two tuned circuits, one involving the primary coil and one involving the secondary coil, and the thing works by putting energy into the primary and then letting the mutual coupling between the coils transfer the energy to the secondary at a much higher voltage.  For further explanations of any technical stuff here, see the Legacy Gen1 manual on the oneTesla website at http://onetesla.com/oneTesla_User_Manual_v1.3.6.pdf.  That manual is for an earlier discontinued model, but does a much better job of explaining how a Tesla coil works than the newer model's manual (http://www.onetesla.com/oneTeslaTS_User_Manual.pdf).

One of many challenges faced by the oneTeslaTS designers was how to squeeze a high-performance Tesla coil, able to put out in excess of 200,000 volts (200 kV), into a small package, less than a foot (30.5 cm) tall.   Most DRSSTCs are at least twice that tall to allow room for the thousands of turns of fine wire wound in a single layer on the long high-voltage secondary coil.  More turns means more voltage, and so obtaining 200 kV from a primary coil excited by 340 VDC (which is all you can get from a full-wave rectifier operated directly from the 120-VAC mains supply in the U. S.) requires lots of turns of wire. 

The way the oneTesla people squeezed all those turns into a coil less than 8 inches (20 cm) long is by using fine wire.  Very fine wire.  No. 38.5, according to the oneTesla manual.  That's about 0.1 mm, less than the diameter of a human hair. 

The only places such fine wire is used routinely that I'm familiar with is in earphones and the old-fashioned "D'Arsonval" analog meter movements.  But such things require close to watchmaker precision and steadiness of hand to build.  Can an average kit-builder deal with wire that fine?  We'll talk about that later.  Suffice it to say that with such fine wire, the designers got enough turns in their secondary to make sparks that rival other units that are twice or three times the physical size of this one.

Another clever trick was to make the primary coil a circuit-board pattern, etched directly into a two-sided board on which the secondary is mounted.  This eliminates any need to hand-wind or insulate a primary coil, which the oneTesla's earlier (Gen1) model required.  It's also probably cheaper than a long hunk of solid wire. 

The driver electronics also has some nice features.  It's controlled (or supposed to be controlled—see below) by an "interrupter board" which is at the other end of a ten-foot fiber-optic cable.  This cable is just a piece of plastic and will not interfere with or pick up the strong electric field coming from the operating Tesla coil.  I learned about this the hard way when I tried to control my previous research Tesla coil with a coaxial copper cable.  The first time I tried it, induced currents fried the electronics, and after that I switched to fiber-optic control, which has caused no problems.

The other problem solved by the driver electronics is tuning.  Like two tuning forks set at the same pitch, the primary coil and the secondary coil have to resonate ("ring") at or near the same frequency for good energy transfer.  And the driver electronics has to send that frequency to the primary or else the driver can't work well and may burn up.  The oneTeslaTS circuit is basically a feedback loop that senses the current in the primary and automatically tunes its frequency to the resonance frequency of the system.  This is a clever solution that avoids a lot of fussy manual tuning that other systems require.

Small, efficient, easy to build—what could go wrong?  Well, quite a few things, actually.  But not until I got past the circuit-board assembly steps, which are by and large straightforward, accurate, and not that hard to do if you know how to solder, how to read resistor color codes, and how to follow instructions.  The interrupter went together without a hitch, but I was left clueless by the manual as to how to operate it.  The main circuit board passed its low-power test, making a satisfying buzz when the interrupter was connected to it.  The next step was completing the secondary coil.  Fortunately, the kit comes with a completely wound secondary, but it's just a coil:  a plastic tube with several thousand turns of wire on it.  It's your job to insulate it better and put end caps on it.

Not So Good News

Five Ways To Break No. 38.5 Wire

I was well aware that the quality of the secondary coil would determine the quality of the unit, so I was careful to follow the instructions about how to treat it.

The secondary is supplied "bare" (only the insulated wire on it) and it's the kit-builder's job to apply enough insulation on the winding itself to keep it from breaking down under the extreme stress of holding off some 200 kV across only a few inches of space.  The manual says to "varnish" the secondary, and shows a picture of one next to spray cans of Krylon Crystal Clear acrylic coating, and Minwax Helmsman spar urethane.  There was some comment in the manual about substituting packing tape, but the important thing is to surround each turn of wire with a dielectric (insulator) that works a lot better than plain air.  I decided to set up a rig (Fig. 1)

Fig. 1.  Coil-varnishing rig.

on my workbench that would slowly rotate the coil about a horizontal axis at a few RPM.  I had a geared DC motor lying around, but an old AC motor from a rotisserie barbecue set would work.  I cut a couple of pieces of cardboard out to make rings that would just fit inside the coil and around the broomstick axle to support the coil on the axle.  The idea is to keep the spray from sliding to the lowest part of the coil and beading up due to gravity.  The manual said to put at least five coats on, and I used five coats of the Minwax Helmsman brand spar urethane, moving the spray evenly back and forth as the coil turned and spacing the coats apart by about an hour.  After letting it dry overnight, the resulting finish was very smooth and thick enough that I have no hesitation in grabbing the middle of the coil and picking the unit up that way.  Don't try this with the bare coil—you'll probably break something and have to order a new one.

The next major operation on the secondary was to assemble the end caps.  The whole point of the end caps is to insulate the two ends of the coil from each other as much as possible, because those two ends are where the 200 kV shows up.  My coil was supplied with several inches of wire at each end, and nothing specific was said in the manual about how much of this wire to use or how long to make it before you assemble the end caps. 

HERE IS THE FIRST SUPPLEMENT I recommend to the manual.  Cut off all the wire from both ends except what you HAVE to use to make the end caps fit.  That is probably about 40 or 50 mm, depending on how you assembled your end caps.

The end caps are stacks of clear insulating plastic laser-cut in various patterns to form a totally enclosed cavity where a ring terminal makes a connection between a bolt going through the middle, and the end of the wire adjacent to the cap.  The manual recommends assembling the caps using HOT GLUE, NOT EPOXY.  I TOTALLY AGREE, because it took me five tries to get the bottom end cap together.  I broke the wire the first four times.  If the layers of the end caps are not almost EXACTLY coaxial (all their outside edges perfectly in line), the end cap will not fit into the inner diameter of the secondary tube and you will have to pry the thing apart.  My wife (who is familiar with hot glue) recommended putting the stack in the freezer for 15 minutes or so to make the hot glue more brittle and easily broken up.  This worked great, but I still had to take a sharp razor or Xacto knife and scrape all the old glue off before lining things up for the new glue. 

As a participant in the oneTesla forum noted, the existing end cap design sort of makes a pair of scissors that tend to cut the wire right where it leaves the top of the tube and runs down between the inside of the tube and the end cap layers to get to the ring terminal, where you have carefully soldered it.  Fig. 2 is a picture (adapted from the oneTesla v. 1.4 manual) of how they want you to assemble an end cap:

Fig. 2.  Manufacturer's recommended way of connecting secondary wire to end cap.

In order to do this successfully, you must (a) solder the fine wire from the coil to the "ring terminal" (I call it a solder lug) that goes around the bolt, (b) while holding the entire assembly out of the top of the coil, but near enough not to stretch or break the wire, finish assembling the top pieces of the end cap together, and (c) slide the entire end cap assembly into the coil WITH THE WIRE ATTACHED, WITHOUT BREAKING THE WIRE. 

This is very hard.

After five tries, I managed this on the bottom terminal, but one failure on the top terminal convinced me to come up with a better way, which leads to this:  HERE IS THE SECOND SUPPLEMENT I recommend to the manual.

Instead of trying to connect the wire to the ring terminal, either save it out before you assemble the end cap, or make your own out of 5-1000ths of an inch thick solderable brass or copper "foil" (available at hobby shops like Hobby Lobby), as shown in Fig. 3:

Fig. 3.  My way of connecting secondary wire to end cap.

I simply soldered the wire at the top end to my hand-made brass ring terminal (cut out with a pair of scissors), and screwed the toroid and the retaining bolt down on top of it all.  (If you save out the ring terminal when you assemble the end cap, you won't have to roll your own.)  This is vastly easier than first connecting the wire to the terminal inside the end cap, and then trying to fit the whole thing, with the wire attached, into the coil.  My way, you can put the end cap on first, and then lay the terminal on top of the bolt and put the toroid on it. 

While the original end cap design does a good job of insulating the wire and the ring terminal all around, there is no need to insulate them from the toroid (on the top) or the ground plane (on the bottom), because the wire is directly connected to those points anyway.  My point is that placing the ring terminal (or a substitute) directly on the end cap surface should not cause any insulation worries, and so far my modifications in this department have not caused any issues.  There is still plenty of end-cap insulation where it counts (between the top and bottom wires), and as long as you keep the wires well away from the hollow part of the inside of the secondary tube, you should not have a problem in this department, as  some unfortunate kit-builders have run into.  But maybe they didn't know what to do with all that extra wire, so they left it dangling inside the secondary tube.  No, no, no.

Blowing the First Fuse

One of the tie points for the power connector is very close to one of the tie points for R5, one of the high-voltage power supply bleeder resistors that discharge the high-voltage filter capacitors to keep you from killing yourself when you take your kit apart to fix it (as you surely will—take it apart, I mean, not kill yourself).  So leaving that resistor off is not an option.  But I was not the only person accidentally to leave a solder bridge between the resistor and one of the AC terminals of the bridge rectifier.  The result was that when I followed the instructions for the first startup and plugged it in, I saw a bright green flash from somewhere on the circuit board and that was that.  The solder bridge had not only blown the fuse, but had vaporized a section of circuit-board trace (vaporized copper glows green, in case you're wondering).  Subsequent waiting and inspection revealed the problem, and a trip to Radio Shack (yes, there's a few left) laid in a supply of 10-amp fast-blow fuses of the right physical dimensions.  Be sure and buy or order at least a dozen of those before you start running your unit.  It's also a good idea to order at least one set of four replacement IGBTs, which oneTesla offers as an option.

This was the last major problem that I can say was definitely attributable to me.  The following issues have more subtle causes that took a few more weeks to resolve.

Eating IGBTs

Here is the way this device is supposed to work.  The interrupter board generates light pulses that occur at a frequency (number of pulses per second) determined by the "frequency" setting of the board.  Just for the record, and for the convenience of everybody who wants to use the "SD" interrupter board, here is THE THIRD SUPPLEMENT I recommend to the manual:  a flow chart for the interrupter board controls.

Controls:  main power (slide switch on left side of enclosure), LCD display illumination (slide switch above LCD display), "select" (*) (middle pushbutton), "up" (^) (top pushbutton), and "down" (v) (bottom pushbutton).  User operations are *, ^, or v.

Condition:  No SD card inserted.
Main power ON (LCD display on or off, doesn't matter): 
>"Checking SD Card"
>"Fixed Mode"
*
>"[Freq:  20Hz ]"

The unit is now producing a 13.6-microsecond pulse at a frequency of 20 Hz.  The pulse frequency can be adjusted upward to a maximum of 1000 Hz (1 kHz) by pressing and holding the ^ button, and lowered down to 1 Hz (the minimum) with the v button.

At any given frequency, the duration ("dwell time") of the pulse can be lengthened by pressing * again:

This moves the pair of brackets [] from the top display line to the bottom display line.  Now, the user can increase the pulse duration ("power") by pressing the ^ or v buttons.  The longer durations are indicated by dark rectangles extending across the bottom of the screen display, up to a maximum of 14 rectangles.

By direct measurement of the pulse duration in my lab, the minimum pulse duration (lowest "power") at all frequencies is 13.6 microseconds.  The maximum pulse duration varies from 48.6 microseconds at 1 Hz to 25.6 microseconds at 1 kHz.  The longer pulses are needed at lower frequencies, apparently to even out the loudness at different rates.  The fact that the maximum pulse duration occurs at the lowest pulse frequency probably explains the rather cryptic statement in the manual that "higher frequencies draw less power from the coil."  This is true if you have run your interrupter power level higher than the minimum. 

If you haven't figured it out by now, the longer the interrupter is HI (on), the more cycles of 260-or-so kHz power the driver will pump into the primary, and the more energy will come out the top of the secondary.  At some point the voltage on the secondary coil will be high enough at the tip of the breakout point (the brass wire supplied with the kit that is shown sticking out horizontally from the toroid), and the energy will start to dissipate into the air.  The discharge at the tip of the breakout point starts as what plasma physicists call "corona":  a burst of tiny fast cold spark-like feathery discharges that absorb some energy and turn it into heat, light, and some chemical reactions (this is where the ozone smell comes from).

In my particular unit, this is what happened the first time I turned the device on (after fixing the solder bridge) in Fig. 4:
 
Fig. 4.  Video of first test of oneTeslaTS.

Although the interrupter was set to low power (short-duration pulse), what I saw in that video did not look like any low power to me.  After unplugging the unit and waiting a suitable time (at least five minutes) for the power capacitors to discharge, I disassembled it and checked the IGBTs.  They had failed, and so had the 10-amp fuse.

 How did I know they failed?  Using the "diode check" function of a DVM, a good IGBT will show about 0.4 V when the collector is negative and the emitter is positive, and open with the opposite polarity.  (See the IGBT data sheet, available online at https://www.fairchildsemi.com/products/discretes/igbts/discrete-igbts/FGA60N65SMD.html for pinouts).  When the device fails, usually a short circuit appears between these two terminals, with either polarity. 

I highly recommend (as the manual does) leaving the full lead length of the IGBTs sticking up from the board when you replace them, because these leads form convenient test points to contact with DVM leads without taking the board stack apart.  Be very careful to place the sil-pads correctly, because any electrical contact between the IGBT's metal case and the heat sink will in all likelihood blow a hole in your heat sink and possibly turn the IGBT into a small bomb, too.  (This is on the user's forum as well, once you register.  You can't see everything until you register.)

I was somewhat mystified by the pretty catastrophic failure of both transistors and the fuse, despite my precautions of starting with the lowest power setting on the interrupter. The way the oneTeslaTS is designed, it is not possible to operate the IGBTs at any voltage lower than 320 V or whatever the voltage-doubler power supply creates from your local power line.  Even with the lowest-duration pulse, this is asking the IGBTs to do a lot if something happens to make the pulse a lot longer than minimum.  And it turns out, that was exactly what was happening.

In one chapter of my book, I discuss a fairly obscure topic:  EMI, which stands for "electromagnetic interference."  EMI is what happens when somebody's cellphone starts bothering an auditorium's PA system, for example.  It is unintentional coupling of a signal to a circuit by means of electric, magnetic, or electromagnetic fields. 

After about four days of trying to get my oneTeslaTS to work at home, I went through two sets of IGBTs, an equal number of fuses, and was no closer to a solution.  I could not understand why even though I set the interrupter to a specific frequency, the sparks would emerge from the breakout point in loud random pops until something busted.  Finally, after modifying the unit to apply less than full voltage to the IGBTs and observing its output with an oscilloscope, I determined the problem:  loss of control by the interrupter.

Loss of Interrupter Control

By using a variable AC supply device called a variable autotransformer (tradename "variac"), I slowly increased the voltage on the IGBTs, slow enough so that when the bad thing happened, they would not be working under the full stress of 340 V and might not immediately fail.  (This was after I connected the driver power supply to a separate constant 120VAC source so that it could operate normally regardless of the IGBT supply voltage.)  Up to about half voltage, I saw that the system was behaving properly.  The interrupter was signaling the driver to allow oscillation for a few cycles, and the secondary was putting out high-voltage pulses exactly in time with the interrupter pulses. Then at about 55% of full voltage, POP!  I saw the scope screen fill with a constant oscillation, much longer than the interrupter would allow it to do if it was still controlling the system.  Somehow, the high electric field from the secondary was getting back into the electronics and keeping the oscillations going even AFTER the interrupter "told" it to quit.

Here in Fig. 5 is a captured pop from a Tektronix TDS2014B scope with its probe picking up an arbitrary-scale electric field (vertical axis) and the timebase in milliseconds (horizontal axis):


Fig. 5.  Electric field (arbitrary units) from Tesla coil during a "pop."  Time axis units are milliseconds.

This explained my problem.  The IGBTs are not capable of operating continuously at 340 V supplying full power to the secondary.  After a few milliseconds of what amounts to a VERY high-power pulse, they heat up so much that they cease to switch efficiently, the field falls, and the EMI loop is broken.  The pulse quits, but the IGBTs are so hot that the system can no longer oscillate normally for a short time.  Then it recovers, runs under interrupter control for a time, the field builds up again, and POP! the cycle repeats.  This is somewhat speculative, but explains the data.

If I had increased the IGBT voltage to full (100%, 340 V) they would undoubtedly have failed again the next time the interrupter lost control.

How is this thing to be prevented?  Commercial and hand-crafted Tesla coils usually have elaborate electrostatic metal shielding around the control electronics.  There is no cheap shortcut to good shielding, so this is why they cost so much.  My first thought was to put on the side panels in hopes that they would provide enough shielding to stop the loss of interrupter control (POPS for short).  But alas! the manual's statement that "installing the cosmetic side panels is optional" turned out to be true.  They are largely cosmetic and did not provide enough shielding to fix my POP problem.

At this point, I turned to the user forum hosted by oneTesla on their website.  I wrote a detailed post describing my problem, and after a few days received a short reply from
Bayley (the "lord protector"), saying it sounded like a grounding problem.  I rechecked grounds and found nothing obviously wrong, so eventually I resolved to replace the entire main board.

This costs about $80.  Fortunately, exchanging main boards is pretty straightforward, mostly a matter of lining up a new set of IGBTs and disassembling and reassembling the board stack. 

I am pleased to report that replacing the main board fixed the popping problem.  The unit runs up to full power at all interrupter frequencies from 1 Hz to 1 kHz without lapsing into the EMI pops.  I did modify the breakout rod to point vertically upward, which moves the area of the discharge farther away from the electronics and may minimize any EMI problems with the unit's own electronics.  The corresponding electric-field waveform from the unit when it's operating properly is shown below in Fig. 6:


Fig. 6.  Electric field (arbitrary units) during one interrupter pulse from proper operation of oneTeslaTS.  Time axis units are microseconds.


Parting Words

If you are careful to avoid solder bridges, do a good job of varnishing the secondary, and make connections to the secondary on the outside of the end caps instead of the way recommended by the manual, you could conceivably have the oneTeslaTS up and running in a week of working on it a couple of hours an evening.  I was not so lucky.   

Although I haven't used the MIDI feature, the interrupter takes a memory card with a MIDI file on it, and so the tunes you can play on the thing are limited only by your stock of MIDI-file tunes (there are five on the oneTesla site to get you going).  Especially if you haven't done this kind of thing before, building this kit will be a learning experience.  But don't start unless you're prepared to devote a lot of time and thought to this project, and a fair amount of filling in the blanks left by the overly brief instruction manual.

REVIEW POLICY NOTICE:  I paid for the above-described kit myself and have no connection with oneTesla or its employees.  I will not review other kits by request, and will turn down any such requests. 

Karl Stephan
Feb. 3, 2016

See my blog on engineering in the news at