BEMF Recovery Circuit

Brandt-Tesla switch finally explained?

By Arend Lammertink, MScEE. April 21st, 2021.

Contact: lamare at gmail dot com

I have recently been looking into some of Ron Brandt's work of which quite a lot can be found here:

His presentation on the 1995 Extraordinary Science conference can be found here:

I was interested in the details of his controller as well as his "perm mag" motor, the combination of which has been claimed to be capable of achieving a COP>1, whereby I was specifically interested in the way he controlled the three phases in his motor. His Perm Mag motor itself is of a design that is very common today: a brushless DC motor with permanent magnets:

Such motors can a/o be found in floppy drives, which would thus be ideal for experimentation with DIY 3-phase controllers:

Even the hall sensors Ron also worked with are included, the three chips that can be seen on the upper part of the motor between the coils.

When we look at Bedini's schoolgirl, the Tesla switch and Ron's perm mag motor, there's two things that stand out:

1) Some kind of "resonance" or oscillation;

2) Capturing BEMF.

In an article in Extraordinary Science (1994) about Ron's work, we find an interesting detail:

"He learned that in a magnetic field, the strength of the field and how fast it collapses determines the amount of energy that can be recovered. The faster the magnetic field collapses, (back EMF) the more energy can be recovered for reuse."

Now let's take a look at how reluctance brushless DC motors (the ones without permanent magnets) are typically controlled, the asymmetric bridge converter:

This is essentially the same circuit as Bedini uses, albeit for three rather than one coil or "phase", with the crucial difference that in this case both the high as well as the low side (of the coil) are being switched, whereby both transistors are opened and closed at the same time. This way, two recovery diodes can be used instead of one and thus the "loop can be closed" and the BEMF can be captured in the capacitor without shortcutting the coil.

This is quite different from how BDC motors with permanent magnets are typically controlled, namely with totempole drivers:

Hereby, the windings are steered in either a star or a delta configuration:

This is also the basic way Ron did it in his perm mag controller according to the schematic, whereby he used IGBT totempole modules and a star configuration for the windings, but there are also considerable snubbler circuits as well as a three phase rectifier bridge intended to capture the BEMF.

The problem with a totempole driver is that the coil gets short cutted and thus the BEMF is wasted rather than recovered:

This is probably why Ron added considerable resistance in the feedline to his IGBT module, his shunt.

I don't know why it is typically done like this with permanent magnet brushless DC motors, it may be that this only applies for low power applications while for high power applications bridges with BEMF recovery are used.

What is interesting, is that with the use of a couple of diodes, we can make it so that a number of capacitors (or batteries) can be discharged in parallel, while being charged (by the BEMF) in series:

Note the similarity with (one half of) Ron's original Tesla switch, from the Extraordinary Science pdf linked above:

Let's now consider the amount of charge stored in two identical capacitors C charged to a voltage V. In series, the total capacitance C_s = C/2, while the total voltage V_s = 2*V. And since Q = C*V, in series we obtain a total charge of:

Q_s = C/2 * 2 * V = CV.

In parallel, the total capacitance C_p = 2C, while the total voltage V_p = V. So, in parallel, we obtain a total charge of:

Q_p = 2C * V = 2CV.

So, it seems that in order to charge two capacitors in parallel, one needs twice as much charge (= current * time) as one needs when charging in series. This is what makes that charging capacitors in series with a constant current goes faster than in parallel with the same current. And we can utilize this by discharging our capacitors in parallel when building up the magnetic field in our inductor, while switching them in series to capture the BEMF with the cap bank circuit shown above.

The advantage of the cap bank circuit is that the switching between series and parallel goes automatically and occurs at the moment the transistors are both switched off (at the same time). It seems that in various incarnations of the Brandt-Tesla switch, the switching between series and parallel is done differently, which might explain why these are hard to tune and only work at specific frequencies thus requiring "resonance".

Either way, I decided to run a few simulations with LTSpice (which can be freely downloaded). You can find the .asc files as well as all the pictures in this article here:

First, I simulated a conventional circuit with one coil ("phase") and one cap:

I plotted the current through the coil, the capacitor and the diode from the power supply as well as the voltage over the capacitor:

The green line is the current trough the coil and as we can see, it charges and discharges in about the same amount of time. The light blue line is the voltage over the capacitor, which shows a sudden jump at the time the transistors are switched off, whereby the current through the cap rapidly reverses in direction.

Now the situation with a cap bank with two capacitors:

And the result:

As we can see, now the coil discharges it's BEMF quite a lot faster, which improves further by adding more caps and diodes:

And the result:

As we can see, this results in the coil discharging (it's BEMF) even faster.


From the simulation results, it appears clear that when utilizing a cap bank with some diodes, it is possible to reduce the time needed to collapse a magnetic field in a coil, be it a single coil as in Bedini's systems, a coil in a brushless DC motor as in Ron's motors or a transformer winding in an inverter such as the Cole-Hackenberger power supply. According to Ron, doing so should maximize the amount of recoverable energy and thus should provide your ticket to "overunity".

Of course, it remains to be seen whether or not this is actually true by experimentation. I intend to experiment with this myself, but may be you are interested in experimenting with this as well. If you do, please do share your findings at one of these forums: