Miniature wireless power demonstrator
I came to know of wireless power at a quite young age, at the same time I first learned about Tesla coils. Naturally, I first understood the concept of capacitively coupled wireless power transfer, and how Tesla planned to do it with Tesla coils. Hence some of my first experiments were actually capacitive coupling, Tesla coil based – see Class E SSTC.
Quickly enough though, I started to realize that much like it is possible to achieve near-field resonant coupling using high alternating voltages and strong electric fields, an analog should be possible with parallel resonance, high currents and strong magnetic fields as well. Tesla coils utilize a high impedance resonant circuit to produce extreme voltages required for capacitive coupling, and this high voltage creates a problem for small scale applications like indoor battery charging (corona and arcing caused by strong electric fields can even cause fires!)
Magnetic coupling is attractive because it allows fairly large amounts of power to be transmitted without need for high voltages. The basic idea is again to have two high-Q resonant circuit, which are now coupled magnetically and are preferred to have as low characteristic impedance as possible. The only question that remained was how to feed RF power at an appropriate frequency into the transmitting tank circuit – and some sort of a self-resonant oscillator looked like a good choice.
Back in time even before I joined 4hv, there was a kind of flyback transformer driver called “Mazzili” or “ZVS” driver, first made by Vladimiro Mazzili, that was popular among HV enthusiasts as an easy way of pumping large mounts of power into TV flyback transformers. This circuit was a variant of a Royer oscillator, such as can typically be found driving LCD screen backlights. I don’t know how exactly I came to an idea to use it for wireless power, but the result turned out working remarkably well and it became very popular among other electronics enthusiasts. This is the schematic of original Mazzili flyback driver:
Obviously, the first modification I had to make to the circuit was to remove the flyback transformer and put an air cored inductor in it’s place. Since I wanted to minimize the characteristic impedance of the tank circuit, this ended up as a single loop of copper pipe. I split the tank capacitor into more parallel capacitors of lower value for higher current rating, and also split the inductor in two going to drain of each mosfet, in order to avoid tapping the center of the copper loop.
The prototype circuit was hacked together with parts I had lying around. The following pictures show an example schematic, and the prototype of the most basic circuit:
The inductors used were of unknown inductance here, consisting of about 25 turns of wire on powdered iron cores from PC power supplies. A single 942C20P15K capacitor was used as a resonant cap, and tended to get rather hot in operation – especially it’s leads! Not surprising, considering it was carrying about 20A or RF current. I used a 12V DC supply for entire circuit – both the Udd and Ugg supplies.
Later I put the circuit onto a PCB and made a better, single-loop receiver coil. Radio frequency chokes, which were also showing significant heating (most likely due to indirect induction heating by transmitting loop’s field) were later replaced by SMT ferrite inductors, and an extra 100 ohm resistor was added in parallel. The tank capacitor was made of 6 parallel 6.8nF 1000V Wima FKP capacitors. The voltage rating is an overkill since capacitors see only up to around 50V, but having the capacitors physically large helps their power dissipation. Both the copper loop and the capacitors get quite hot in this application!
I haven’t bothered measuring efficiency at that point yet, because I knew it was rather low due to high power dissipation on gate pullup resistors, and the fact that output was AC making it difficult to accurately measure power. The circuit operated at about 1.5Mhz.
The initial prototypes of this circuit had a problem: Supply voltage rising too slowly could cause the circuit to latch-up in a stable state where one mosfet is fully on, and another fully off – with bad consequences if no protection is utilized! I thought to build in some sort of electronic threshold circuit that would enable the gate supply only after the input voltage has risen to a certain level, but this started to seem too complex for the spirit of this circuit – so I decided to use a relay instead to perform the same task, which was a simple and effective solution.
Some additional info on the components used that isn’t shown in schematic:
- R1, R2, R6, R7 – 100 ohm, 2W
- C1-C8: 6.8nF, 1000V WIMA MKP capacitor
- L1, L2: 100uH, 3A inductors
- R3, R4, R5 – 1k, 0.25W
- C11, C12 – 100nF, 50V
- K1: 12V coil, 5A relay
- PAD 3 and PAD 4 on schematic are connections for the transmitting loop
- JP1 and JP2 are jumpers, designed in to allow insertion of a step-up transformer between the driver and the resonant tank circuit. Only tried it once though, with a 1:2 autotransformer, and felt I’m already having too much heating in the capacitors to push the power further. The supply voltage was limited to about 15V due to relay and MOSFET gate rating, but this seemed enough to perform fun demonstrations. The following picture shows the PCB layout:
I’ve been asked countless times why have I just used just one turn coil instead of several turns, unlike the original MIT witricity experiment which used several turns of litz wire for their coils. The answer is again in the characteristic impedance, which I wanted to keep as low as possible (because it is proportional to square root of L) in order to push as much power as I can with the fixed input voltage I had available. If I used more turns, the amount of power I could transmit over a certain distance would drop!
MIT team, on the other hand, had no choice but to use several turns, in order to prevent their resonant frequency from going into Ghz range and their system turning into an efficient radiator (something we absolutely don’t want with wireless energy systems!) This is because, unlike me, they didn’t use a resonant capacitor, but only relied on interturn capacitance much like the case is with Tesla coils. High resonator voltage probably didn’t matter for them at that point, since they fed the transmitting resonator via transformer action, from a separate vacuum tube oscillator. Most 4hv members at the time didn’t see anything more extraordinary in their system, other than a pair of quite efficient Tesla coils with relatively low number of turn and high operating frequency.
On the other hand, my variant is probably more like an induction heater than a Tesla coil, considering it has a resonant tank capacitor and really low characteristic impedance. I’ve actually built a small induction heater based on a design very similar to this one!
And finally the video which has attained a cult following on youtube. Countless people have been mailing me so far for schematics, help and plans for this circuit and many have succeeded replicating it on their own. Please be reasonable, however, that it may still be difficult to find exactly the same parts like I used in your country, and that some PCB modification is likely going to be necessary! Please read the FAQ and look at Eagle files in the 4hv thread if you want to build this circuit.
This demonstrator was ultimately donated to Nikola Tesla museum, in his hometown of Smiljan where it resides till this day.
After completing the demo circuit, I started thinking about how could I improve it’s efficiency. One thing that would certainly be of great help is to get rid of the pullup resistors. I’ve experimented with active pullups a but, but for some reason they made the circuit even more unstable and prone to latch-up.
Then one day the weirdest idea struck me – I thought, if I could pick up a bit of the main coil’s magnetic field with one extra coil, and feed this back to mosfet gates while increasing the resistance of the pull-up resistors. The idea somehow worked extremely well and I was able to reduce no-load current to about 200mA from previous 700mA+!
This time I’ve also rectified the output, by using another Royer circuit as a synchronous rectifier (yes, it works in reverse too, and bidirectional power transmission can be achieved this way!) I’ve used a motor as a load that drew about 1 ampere at 15V. At 10cm distance, the efficiency was just about 50% – clearly there is still a lot work to be done before this simple circuit could find an useful application!
Of course, I didn’t stop there: my research on these demonstration circuits were later very important for my BSc thesis wireless mouse project. During that project I did large amounts of research on Royer oscillator, and learned some important lessens about feedback oscillators in general (Which most solid state Tesla coils are too!)
The most important one has to do with formerly mentioned latch-up problem, which has long been a mystery of this circuit, and has to do with system stability. An oscillator is one rare example of systems which we want to be unstable in it’s normal mode of operation; that is not to satisfy Nyquist criterion!
Phase criterion is pretty definitely violated once we take positive feedback from the output; it now all depends on whether the circuit will have enough gain to oscillate. And trouble with nonlinear circuits is, that their gain may not be the same over entire operating range. In case of MOSFETs, it is pretty clear that their gain will be about zero if they are fully on, or fully off! If the circuit is allowed to stabilize in a condition where one MOSFET is permanently on and another off, it won’t be able to recover because there is not enough small signal gain to start the oscillation. MOSFET that is on then just shorts the power supply until either something blows or protection devices trip.
This is why the original circuit required a fast rising supply voltage to start – when the voltage is suddenly applied, the inductor current rises relatively slowly, and at some point MOSFETs will pass through their linear region (as the gate voltage rises) where they have high gain and the oscillations will start. If the voltage rises very slowly compared to inductor time constant, the gate voltage may never rise enough to start the oscillations and the circuit will latch up.
The solution to this problem is to bias the MOSFET gates within linear region in steady state (and not to full +12V as it was done by pullup resistors!). However, since MOSFET’s are designed for fast and efficient switching, their linear region is very narrow – and apart from that, it varies widely with temperature, having a negative temperature coefficient.
If we tried to bias a MOSFET with a simple resistive divider, it’s pretty certain to turn further and further on, as it heats up, until a latch up occurs. This makes it much more difficult to bias MOSFET oscillators, than ones based on vacuum tubes or bipolar transistors. I’m certainthis explains quite a bit of failures people experience when trying to construct a Hartley or Armstrong oscillator using MOSFETs!
I contemplated two possible solutions to this problem:
- Build a temperature-sensing bias circuit that would compensate the effect of temperature
- Use a microcontroller to generate a PWM signal which is then filtered to produce bias voltage. The duty cycle is slowly increased until the system starts to oscillate, and the current is monitored by an ADC to detect any problems.
Naturally, I first started contemplating the simpler solution, without the mcu – I tried thermally coupling 4-5 diodes to the MOSFET package, and forward biasing them by a resistor to produce a bias voltage which is dependent on temperature, due to NTC nature of diode forward voltage drop. I tested this idea and it worked well on a small Royer circuit, but mounting the diodes onto MOSFET proved to be rather troublesome. A NTC or PTC thermistor could be used instead of the diodes as well, also provided that it can be well thermally coupled (this can be extra difficult with SMT MOSFETs!)
In the most recent times I’ve started to move all my power electronics towards digital control, and I think this is the real future that should be followed. The mcu can directly adjust bias drain currents for both MOSFET devices if necesary without need for temperature measurement, and provides the extremely valuable over-current protection.
Stay tuned for further developments in this area!
Links and references
 4hv thread, with an extensive faq and a number of Eagle files regarding this circuit
 an example replica circuit
 MIT experiemnt