A 65 watt single 6146B
Tesla Plasma Tweeter
Henry Hurrass KB7OCY
I was originally inspired to build this plasma tweeter by Carl A. Willis KA4KIG. His article and schematic can be seen here. You will notice that the input stages are exactly the same with no changes. Where this one differs is in the driver output and final stages. I felt there were enough changes that they should be described in detail.
Watch the youtube video here.
(Click on images for larger picture)
Schematic can also be downloaded in printer friendly PDF format here.
Bottom View of Chassis
GENERAL CIRCUIT DESCRIPTION AND ASSEMBLY
Audio Input Stage
A low level audio signal is applied to the input jack and passes through a high pass filter consisting of C2 through C4, R2 and R3. R1 provides a low input impedance which helps to reduce the amount of noise picked up by the input stage and C1 helps bypass RF to ground. AUDIO GAIN CONTROL R4 adjusts the signal level into V1a's first audio stage which is self biased by combination of C7 and R6. V1a is cascaded to V1b for additional gain. Note the use of ceramic capacitors and carbon resistors. I recommend always using carbon, metal film or other non-inductive resistors where signals are present; this is important.
Audio Driver Stage
I used the same driver tube as the original author, as I had a couple of new 6CL6 tubes on hand. These tubes may be somewhat scarce to obtain. If I had to do it over again, I would most likely choose a 6BQ5, as these, like the 12AU7,s are readily available at most guitar shops. Note that there are four 33K / 2watt metal film resistors in parallel (~8250 ohms) for the plate resistor. Here's where my circuit required changes. The approximate voltage on the plate with no signal was originally in the 160~180 volt range. Also the low voltage on my power supply was 385 volts. Just the plate current of the 6CL6 alone at idle causes dissipation of over 6 watts at idle. Keep in mind that the driver tube also must supply bias voltage (and current) to the final 6146B screen grid as well. The additional burden of the final caused the plate voltage on the 6CL6 to drop to less than 70 volts. The plate resistor dissipation jumped to over 12 watts, causing the solder on the resistor assembly to instantly melt! Not good.
Two things needed to be done: Low voltage plate supply needed to be reduced and better control of the screen drive / screen bias voltage was necessary. I'll explain the easy and simple "power supply voltage reducer" modification later. First, let's discuss a few things about screen bias. The screen voltage directly affects the plate current through the tube. The ideal voltage for the driver tube seems to be in the neighborhood of 140 volts or so (or about half the plate supply). Problem is: With 140 volts biasing the final's screen, the final would draw 120mA or more of plate current. Assuming a plate voltage of 675 volts:
E x I = Watts (DC) or 675 volts x 0.120A = 81 Watts plate dissipation
The 6146B is only rated at 65 watts, so we are exceeding the tubes rating by 16 watts; and that's with no signal! The tube plate will glow red and it's life will very likely be shortened.
So if we reduced the screen voltage, the plate current would drop accordingly. Trouble is, this becomes a rough balancing act that would tend to drive one nuts.
Final Screen Biasing Solution using a MOSFET Follower
Referring to the schematic on the left, we see the driver tube's plate dropping resistor R12. With a plate supply voltage of 330 volts and a measured plate voltage of 140 volts, we get a dissipation of 4.37 watts with no signal across R12. R12 still gets pretty warm, but with nearly two times the rating (I recommend a minimum of 2X ~3X rating on any resistor), we really don't have anything to worry about.
A MOSFET follower (Q1) is used to isolate the output of 6CL6 driver tube from the screen input of the 6146B final. The bias is set with R14 and R15 and can be adjusted from approximately 70 to 140 volts. I set my bias to about 110 volts (95mA plate current) for about 64 watts of final plate dissipation. The beauty of the MOSFET follower is the extremely high input impedance and very low output impedance. The audio signal appears across R17, which consists of two 10K / 3W metal film resistors in series. Since the dissipation across Q1 is quite high, Q1 must be mounted on a substantial heatsink. I did fry my first MOSFET, but I think it may have been a combination of several factors, including RF feedback and not enough heatsink. I therefore recommend ferrite beads on the Source and Drain leads, and, instead of using a RF choke in series with the final screen grid, I used an iron-core series choke from an old radio with a 50uf audio bypass capacitor across the choke. Using a plain old RF choke in this configuration caused severe RF feedback and interference, as well as a blown MOSFET; so I don't recommend this.
Plate Current Metering and Screen Bias Adjustment
Screen bias adjustment is mounted on top of the chassis next to the plate current meter. This is a 0~1mA full scale meter. Calibration resistors for 200mA full scale consists of shunt resistor R18 and series resistor R19. Bypass capacitor C16 actually consists of several ceramic capacitors mounted as close as possible around the tube socket. These capacitors keep RF out of the metering leads as they exit the tube socket. Shunt resistor R18 (two 1 ohm resistors in parallel) is also mounted as close to the tube socket and ground as possible. Location of series resistor R19 is not that critical. It's value is about 8.2 ohms. An additional RF bypass capacitor is placed across the meter terminals. Of course you won't have to worry at all about the metering if your power supply has a plate meter installed. Just connect the cathode directly to ground.
Bypass capacitors shown around 6146B tube socket. Meter shunt resistors can also be seen. Grid input resistors consists of two 10K / 3watt film resistors in series.
Photo on right shows meter bypass capacitor and 2SK1050 power mosfet follower.
Plate Tuning Capacitor
It seemed to me that the 6146B final was driving into a load that was the wrong impedance. I was only getting about 3mm corona jet from the stinger using the pheonolic secondary. The audio output was disappointingly weak as well. No other change caused more of an improvement than adding a plate tuning capacitor to the circuit. Even untuned the secondary output instantly doubled!
Another unique characteristic observed is the ability to tune out most of the audio distortion as well. I highly recommend this addition! The plate tuning capacitor consists of C21 and C22. C21 acts as a voltage divider and C22 is the air variable plate tuning capacitor. C21 is a 100pf / 500volt mica cap. Simply adjust the tuning capacitor for the best sound. You may have to re-adjust the screen bias as the tube will draw more plate current as it begins to conduct more in it's "linear" range and the audio improves quite noticeably. RFC1 is a parasitic choke consisting of four turns of wire on a 47ohm / 1watt carbon resistor. The resistor serves merely as a "coil form". Plate choke was recovered from an old ham radio set. Choke needs to be of sufficient size to handle the power and "band of operation"...speaking of the secondary Fo, of course. I used plenty of RF bypass capacitance including ceramic and paper caps as close to the base of the plate choke as possible. Remember, the entire RF signal appears across the plate choke, so it has to be a good one!
The grid feedback consists merely of C17, R20 and a feedback antenna. This is a conventional arrangement that has been around for many years. The feedback antenna is made with a modified banana jack slid over a piece of brass brazing rod for vertical adjustment. The bottom piece of brazing rod is connected to another banana jack which also doubles as a concentric horizontal adjustment. The antenna picks up the current produced by self resonant secondary and is coupled to the grid via coupling capacitor C17. R20 consists of two 10K / 3watt film resistors in series and serves as bias for the control grid. I measured about -40 volts on the grid. I tried a couple of different secondaries and the adjustment didn't change much; but it's sure handy to have!
The output stinger is a tungsten welding rod typically used in TIG welding torches. The end is ground to a sharp point. This material seems to last a very long time. It does get very hot.
The photo on the right shows a blocking capacitor being used in series with the secondary. I don't think the size is all that important, but it does serve to keep the plate voltage out of the secondary; very important for safety considerations. Speaking of safety, be sure not to place yourself or others anywhere near the plate circuit. The high voltage present here is very dangerous and can be lethal!
I made a number of different secondaries for experimental purposes to find out what effect different coil form materials would make, and what effects different frequencies would have with the audio modulated Tesla Coil Tweeter. The coil forms are all about 1 1/4" (30mm) diameter. The first coil was wound with three inches of #24 AWG on a pheonolic form. The Fo is about 6Mhz with an average 5mm brush output. I noticed the form got pretty hot during operation, so I tried an experiment. I placed two forms with a glass of water in the microwave oven for one minute. One form was pheonolic and the other was fiberglas. The pheonolic got hot to the touch and the fiberglas was barely warm. I figured the fiberglas was a lot less lossy, so I wound the fiberglas form with the same number of turns of #24 AWG (same Fo ~ 6mhz). The results surprised me: About 50% more spark length (now at ~ 8mm brush discharge!) I made a third secondary with 3" of #20 AWG with an Fo of about 9mhz on a fiberglas form. The discharge looks more like a flame with this one and reaches over 3/4" (~20mm) on audio peaks. The audio sounds different with the 9mhz coil. Seems like it's more mellow and not as harsh as the 6mhz coils. Most people like the sound of the 9mhz coil better. The output tube plate does start to glow dimly at rated input power with the 9mhz coil as well. Definitely runs hotter. When winding the secondary, wind it tight and don't plan on adding a coat of varnish or other insulation to the finished winding. It may run so hot during operation as to blister the coating.
One very important thing to remember when building a project like this is proper grounding. Let me clarify here; I won't be talking about earth ground for the moment, I'd like to talk about "Star Grounding" instead. Star grounding is a technique for reducing or eliminating ground loops that can cause all kinds of problems like "hum" and "feedback", "instrumentation error", etc. These problems can be a nightmare to solve; particularly if you "designed them into your layout". It can be a very difficult task for even the best engineer to properly layout components in a fashion as to observe all current flows which could "combine" with one another inadvertently to produce grounding problems. Star grounding differs from the conventional ground plane in that all currents flowing through a component flow back to the power supply and only the signal is passed on the the next stage. It's actually quite simple to plan and execute; if you follow a few simple rules. Take a look at the input jack in the photo above. Notice that it's ground is NOT connected directly to the chassis. Instead, the ground is brought back to the first stage of amplification and "chassis grounded" there and only there along with all the cathode resistors and capacitors of ALL the tubes. A little "fudging" so to speak was done at the base of the final, for RF bypass purposes; but even there the ground is so close to the "chassis star ground" that circulating currents would be minimal. On a bigger chassis with more tubes this technique may not be 100% practical. I that case you would make a "ground buss" and connect only one connection from the "ground buss" to the chassis....where the signal currents won't combine with power supply currents. Happy Grounding!
This project uses a quarter-wave Tesla resonator which requires a good connection to earth ground. I provided a ground stud on the back of the chassis and carry a short jumper to tie the chassis to ground (or in the case of my steel work bench; I'll clip it to the bench). I changed my power cord on my power supply from a 2-wire to a 3-wire grounded cord. This considerably reduced the RF zaps I was getting from touching the chassis. The RF field seem to be quite strong as it occasionally jams my little CD player, so you may have to move the player further from the field.
This project was a lot of fun to build and people get very intrigued watching it. If I were to do it again, I would probably us a pair of 6146B's for the simple reason that when it really starts going good, I run out of power. I had no way of knowing this originally ,or I would have left a little more room on the chassis. As it is now; it's pretty compact.
For my project I happened to have an old Ham Radio power supply I could use to power my tweeter. Of course, nothing's perfect. The filament was rewired from 12V to 6V. That was easy. The main problem was an excessively high low voltage B+ supply. This seems to happen more often than less. I expected about 300V but got nearly 400V. There is a simple solution. You need to build a "Voltage Reducer". The circuit is simple a MOSFET follower with the correct zener diode on the gate. The zener itself couldn't handle the load current itself without being excessively large and expensive, so why not use a cheap low power zener diode and let a cheap $2 power mosfet do all the work? You just need to bias the zener for much less than it's max current and let the high gate impedance of the mosfet regulate the heavier source current.
Photo on left shows how simple it is to build a voltage reducer. It consists of a zener diode in series with a 2K resistor connected from drain to source. A 270 ohm resistor is connected from reference to gate. I used two 27V zeners on mine and ended up with a 60V reduction in plate voltage (from 385V to 335V). The tubes ran much cooler thanks to this trick. Mosfet should be mounted on a heatsink.