Skip to content

PL509/PL519 half bridge base driven coil


Base driving a Tesla coil means feeding energy from a high frequency AC voltage source directly into base of the resonator, instead of having it connected to ground and energy being driven in by magnetic coupling. Due to high impedance of spark-loaded resonator, quite high voltages are requires – several tens of kilovolts in order to achieve typical spark lengths.

A great benefit would be complete absence of primary coil, removing any possibility of primary-secondary arcovers, as well as absence of generally bulky and expensive primary resonant capacitor in dual-resonant coils. The disadvantage is that it is quite difficult to produce high AC voltages in frequencies of hundreds of kilohertz – attempts to du it using a HV transformer have so far pretty much failed due to excessive parasitic inductances and capacitances in the transformer limiting the attainable output voltage.

I had a following idea: Vacuum tubes, especially pulsed ones, already operate at voltages suitable for base feed, and can switch at very high speeds.

Usual vacuum tube coils were generally based on an Armstrong oscillator with a parallel resonant primary circuit – the naturally high output impedance of the tube amplifier would be first converted down by the parallel LC tank, and then back up by transformer action – a feat I considered completely unnecessary.

There was also another variant – a base feed circuit that was exclusively used for plasma speakers (tweeters) that produced small plasma flames in Mhz range. The feedback was taken by an antenna that picked up a portion of Tesla coil’s electric field and drove control grid of a tetrode. The amplifier was still single ended, with a single choke towards the high voltage DC supply.

While this is a viable base feed method known for many years, nobody ever seemed to have tried upgrading it into a Tesla coil of more serious size. I could see two main problems with it:

  • The single ended amplifier requires a loading capacitor in parallel with the tube, that usually needed to be variable and carefully adjusted according to operating frequency and load impedance, in order to attain zero-voltage switching and good efficiency. While this may be easy to attain for a small plasma tweeter with a variable capacitor from an old radio, variable capacitors for voltages in tens of kilovolts are rare, cumbersome, and expensive. Also, the load impedance and resonant frequency of the coil changes during operation – not only based on power level or spark length, but also dymanically during spark growth, as seen in QCW coils!
  • I considered antenna feedback such as used in plasma tweeter circuits unsuitable for a high power coil, because it provides drive that is pretty much sine-wave and may cause slow switching in the tube, leading to losses. The grid could be overdriven into deep saturation, but on a high power coil this increases the danger of melting the grid! On the other hand, I also wanted to be able to control the coil by a digital circuit and do all sorts of modulation of it, including QCW operation. This required a driver circuit of some sort that would provide a square wave to the tube – such as a gate drive transformer, or a floating fiber-optics controlled driver

My idea that solves both of these problems was certainly something that world has never before seen – a half-bridge amplifier of vacuum tubes, with the Tesla coil connected to it’s midpoint. A half-bridge requires no additional resonant capacitors making it a truly broadband topology. An obvious hurdle here is how to provide heater power and drive signal to the tube on high side of the bridge, that floats at kilovolts over ground potential!

Obviously, this problem has made many vacuum tube circuit designers never even attempt such a topology – but, for me, this was a great opportunity to make use of my wireless power circuits! The voltage used to power the filaments could easily be multiplied (because it’s high frequency AC) in order to provide positive and negative voltages required by the grids. The drive signal could be provided via several methods:

  • Grid drive transformer
  • Fiber optic link
  • Nearfield communication, perhaps using the same field that provides power!

I chose the grid drive transformer for my first experiment because of it’s simplicity – however, they are difficult to insulate at high voltages and frequencies. Since this was a highly speculative experiment, I wanted to keep the number of variables down.

Then there comes the choice of tubes. While this topology could be used with some truly massive tubes, I decided to settle for small and easily available tubes – PL509 TV sweep tubes. This particular tube has several advantages for this application:

  • Being designed as switching tubes, they have remarkably low ON conduction-voltage drop – as low as 30 volts at their full rated current!
  • They are beam tetrodes, making them easy enough to drive over a gate drive transformer.
  • They are designed to withstand high peak flyback voltages, as high as 7kV, for about 20% of their duty cycle. But since tube ratings are more like guidelines, I thought this could be extrapolated to perhaps 5kV at 50% duty cycle – exactly what I need for a half bridge. And 5kV (2.5kV amplitude, actually) and 500mA is already a rather fine power level for a smaller sized Tesla coil!

I discussed this idea a lot with Steve Conner before I put it into realization. While he liked the idea, he was skeptical at first that I could really attain significant amounts of power with this kind of amplifier. But after a while, he found a very interesting article about a radio amplifier that used two PL519’s to produce 1kW of continuous power in nonlinear mode of operation:

This amplifier had a switching power supply that was modulated by the input signal, and would transit between linear class AB operation and saturated class C at high power levels. For an amplifier used to drive a Tesla coil, there is little benefit from linear operation: it was best to keep the tubes switching as quickly and efficiently as possible all the time. But nevertheless this article has shown an example that substantial amounts of power can indeed be produced by low-conduction loss TV output tubes! Still, some unknowns remained – for example, Steve thought that my grid drive transformers could blow up, a reasonable fear considering many kilovolts of HF AC present. However, they didn’t, and I managed to get some interesting results from my experiment!


The whole circuit is actually very much like a MOSFET half-bridge, with the MOSFET devices being replaced by tubes. I decided to put everything onto an single PCB that is divided into three sections: one of the houses a low voltage Royer oscillator dedicated to wireless energy transfer. The other two sections are tube sections, each having it’s own support circuitry. Since these sections are floating on 5kV potential, they needed to be spaced considerably on the PCB.

Each tube section also had it’s separate drive transformer, for which I’ve had to go through some trial and error in order to construct them properly. The transformers needed to be generously insulated and yet tightly coupled to minimize ringing on the drive signal. A schematic of a single tube module is shown below:

One of two PL504 modules used in a half-bridge. Pads 1 and 2 connect to the wireless power receiving coil, while pads 3 and 4 connect to drive transformer secondary.

One of two PL504 modules used in a half-bridge. Pads 1 and 2 connect to the wireless power receiving coil, while pads 3 and 4 connect to drive transformer secondary.

The wireless power scheme is quite simple, using the Royer oscillator identical to those in my later demo circuit. The high frequency AC from the receiving tank circuit is branched towards two rectifiers: the first is a full wave rectifier of fast diodes, followed by a filter capacitor that provides around 27V DC for the tube heaters. The exact voltage is adjusted by varying the Royer oscillator supply voltage. Another is a Cockroft-Walton multiplier made from 1uF ceramic capacitors and fast diodes, that has both a positive and a negative side. The positive voltage is used to power the screen grid, and the negative provides bias for the control grid (without it, the tetrode would always be turned on!). The screen has a simple linear voltage regulator made around a MOSFET to allow the voltage to be varied.

Royer oscillator used for wireless power.

Royer oscillator used for wireless power.

Royer oscillator was kept as simple as possible. The inductors were picked up from scrap and I don’t know their exact value, but they look like few tens of microhenries. A feedback coil was used to minimize the losses on gate pullup resistors. Two parallel transmitting LC circuits are driven from the same oscillator: resonant frequency doesn’t change no matter how many parallel LC’s are added! 2-turn, ~7cm loops of ~3mm diam. copper wire were used along with 56nF capacitors, resulting in a resonant frequency of around 800kHz.

The whole PCB. Some tube pins are deliberately left unconnected, in order to facilitate swaping the PL509's for PL519's which have similar pinout

The whole PCB. Some tube pins are deliberately left unconnected, in order to facilitate swapping the PL509’s for PL519’s which have similar pinout

Immediately after first power up, I ran into a problem: the UF4007 diodes, which I used for high frequency rectification, quickly started to overheat and blow up, even under a slight load. Apparently, 800kHz was still too much for them despite their speed! I had to replace every diode with two series 1N5819 schottky diodes – they handled this frequency well, but two in series were needed due to their 40V voltage rating being too low for the CW multiplier use (diodes see 2x peak input voltage there!) Now, some real life pictures:

The finished product

The finished product

Closeup of the Royer oscillator section

Closeup of the Royer oscillator section

A closeup of the grid drive transformers

A closeup of the grid drive transformers. Freewheeling diode strings can be seen in this picture, which aren’t shown on the schematic.

A few more comments about the drive transformers: the tetrodes fully conduct when their control grids are driven to about 0V in respect to cathode, and need -50V or so to turn off. The voltage multiplier provides about -80V to assure this. This means that to turn the tetrode fully ON, the transformer needs to provide at least +- 80V. I used a signal generator that provided about 20Vpp to drive the transformers, and this implied a turn ratio of around 1:4. I used slightly more, and a series resistor on the grid to limit the current during ON periods. What I first failed to realize, however, was that rectifying action of the control grid (when it is forward-biased by drive signal) tended to drive down my bias voltage lower and lower, until the tube wouldn’t turn on – this had to be solved by adding a “grid leak” resistor that loads down the bias supply enough to keep it from drooping. Note that these resistors aren’t shown in the upper schematic, though both are present on the PCB.

The transformers needed to be heavily insulated to withstand several kilovolts of HF AC present. I managed this by making the primary winging from RG174 coax cable that was pulled through thick PVC sheath and an extra layer of heatshrink tubing over it. The core of the cable was used as a primary, and the shield was grounded on one side for safety. In case a breakdown occurred, it would hit the grounded shield first and prevent any low voltage circuitry on the drive side from being raised to dangerous potential.

Voltage multiplier and linear regulator

Voltage multiplier and linear regulator

Orange glow in both tubes indicate wireless powered heater success!

Orange glow in both tubes indicate wireless powered heater success!


Now there came time to do some powered testing. I used a single MOT (Microwave oven transformer) level shifted with a microwave oven diode and cap combo – hopefully resulting in 5-6kV peak supply voltage for the amplifier. I used my signal generator to provide a drive signal for this coil, which meant that tuning was open-loop with no feedback involved. A small 50mm coil was used in this experiment, with considerable topload because this seemed to benefit base fed coils by reducing their base impedance, so more power could be fed in for a limited supply voltage. This setup managed to produce some 10-15 cm long sparks!

After some thinkering I realized that my current level shifter may not be up to the task, and that the voltage may be sagging a lot with only 1uF capacitor. Hence I decided to soup it up the following way: one MOT was full wave rectified and filtered by 47uF bank of electrolytic capacitors, and another MOT was added in series to produce a level-shifted waveform. I liked this approach because the inductance of the extra MOT would limit the current in case anything goes horribly wrong, so the things don’t get exploded to bits by several hundred joules stored in the capacitors. I also brought in a big 3-phase variac to control my supply voltage. Due to size and weight of this variac the whole setup had to be moved into the garage. The 3-phase outlet was on one side of the garage, and single phase on another, resulting in a quite haphazard setup that stretched all across the garage. I had to walk in a big circle around it in order to control the variac, or adjust the signal generator frequency. I decided to give it all I can this time, and turned up the variac until the MOT’s were buzzing loudly from saturation, and the voltage went over 6kV in this case. The power consumption went well over 1kW and some nice 20cm+ arcs were produced, but I also got an unpleasant surprise of a tube arc occurring during run. Interestingly, the arc didn’t leave any permanent damage to the tube, and the coil continued running after it happened. Satisfied with the result, I dismantled this risky setup after the run.

Note that base-fed Tesla coil arcs are far more dangerous than with typical VTTC’s, because of extremely lethal voltages and currents available. Despite the coil has been decoupled from the amplifier by a DC block capacitor, but that’s not something I’d trust my life onto. I was extremely nervous during running this setup, and had to watch my every step not to trip and come in contact with lethal voltages. It certainly isn’t something I’d recommend to other experimenters. I decided I have to make it a lot better for the next time, and enable myself to control the coil from much greater distance , and have all high voltage circuits properly enclosed, before continuing to play with it more and taking some proper measurements.

Until then, I can quite clearly conclude that this kind of an amplifier is a sound approach, and can indeed be used to push lots of power efficiently with rather small vacuum tubes. There was no observable plate reddening even with prolonged runs at power levels over 1kW, and by observing the power levels at which this occurred during DC load tests, this means roughly less than ~30W of dissipation per tube, and, counting other losses, perhaps less than 100W total. This gives overall amplifier efficiency of over 90% which is an extremely good result for a tube amplifier!

However, there were some significant pitfalls in this design as well. The major one was that filament voltage would sag significantly whenever drive signal was applied and the screen would start to draw current. This was because wireless power system had no any real regulation built in, and is something I should have anticipated with this project. I hoped that the sag would be insignificant with the transmitting and receiving coils placed so close, but I was wrong – I got almost 10 volts of filament voltage drop, and the filaments would turn very dim. I had to increase the filament voltage to over 40V in no-load condition which the tubes handled for few experimental runs, but is not something to be continuously practiced.

Another pitfall were the drive transformers. In the end I ended up with a situation where I had to use either large value series resistors that slowed down the switching, or low value grid leak resistors that dissipated large amounts of power. Trying to find the best tradeoff resulted in some charred resistors on the PCB, and I still felt the drive was less than satisfying. The insulation on the transformers wasn’t great either – In one occasion I managed to tune the coil onto a wrong mode and had corona show up on the transformers! I abhorred the situation where the transformers might break down, and safety grounding wires for whatever reason get disconnected – this could send 6kV to the signal generator which I might be touching at the moment! For these reasons I decided to phase out the transformers, and retire the current version of this circuit. I decided to concentrate on fiber optics and dedicated floating grid drivers in future.

I also have two new PL519 tubes I could use to enhance this project, but I decided to withhold that for now, since this approach really makes much more sense for larger tubes. For this reason I decided to step it up with some 715C pulse tetrodes. This time a stable and reliable power supply will be required, as well as a powerful high-voltage-swing grid driver which will be investigated in my pulse tetrode driver pages. One day, I would like to apply this idea to some of the very largest of tubes, driving a massive Tesla coil in megawatt power range and produce some QCW Sword streamers of unseen lengths!

Links and references

[1] “Quaggi” radio amplifier served as a part of inspiration for this project


Leave a Comment

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: